@arp-protocol/cli 0.0.1 → 0.0.3

package/dist/hook.js

@@ -14,9 +14,496 @@ function appendOwn(path, text) {
 }
 
 // ../hook/src/hook.ts
-import { existsSync as existsSync2, lstatSync as lstatSync2, mkdirSync as mkdirSync3, readFileSync as readFileSync3, writeFileSync as writeFileSync2 } from "node:fs";
+import { existsSync as existsSync3, lstatSync as lstatSync2, mkdirSync as mkdirSync4, readFileSync as readFileSync4, writeFileSync as writeFileSync2 } from "node:fs";
 import { homedir as homedir3 } from "node:os";
-import { dirname as dirname3, join as join4 } from "node:path";
+import { dirname as dirname4, join as join5 } from "node:path";
+// ../../packages/graphql/dist/http.js
+var DEFAULT_TIMEOUT_MS = 1e4;
+var DEFAULT_MAX_BODY_BYTES = 8 * 1024 * 1024;
+
+class HttpError extends Error {
+  kind;
+  timeoutMs;
+  constructor(kind, message, timeoutMs, cause) {
+    super(message, cause === undefined ? undefined : { cause });
+    this.name = "HttpError";
+    this.kind = kind;
+    this.timeoutMs = timeoutMs;
+  }
+}
+async function readBodyText(response, options = {}) {
+  const maxBytes = options.maxBytes ?? DEFAULT_MAX_BODY_BYTES;
+  const timeoutMs = options.timeoutMs ?? 0;
+  const body = response.body;
+  if (body === null)
+    return "";
+  const reader = body.getReader();
+  const signal = options.signal;
+  let onAbort = null;
+  const aborted = signal === undefined ? null : new Promise((_, reject) => {
+    const fire = () => reject(new HttpError("timeout", `timed out after ${timeoutMs}ms reading the body`, timeoutMs));
+    onAbort = fire;
+    if (signal.aborted)
+      fire();
+    else
+      signal.addEventListener("abort", fire, { once: true });
+  });
+  aborted?.catch(() => {
+    return;
+  });
+  const decoder = new TextDecoder;
+  let total = 0;
+  let text = "";
+  try {
+    for (;; ) {
+      const chunk = await (aborted === null ? reader.read() : Promise.race([reader.read(), aborted]));
+      if (chunk.done)
+        break;
+      total += chunk.value.byteLength;
+      if (total > maxBytes) {
+        throw new HttpError("too-large", `body exceeds ${maxBytes} bytes`, timeoutMs);
+      }
+      text += decoder.decode(chunk.value, { stream: true });
+    }
+    return text + decoder.decode();
+  } catch (error) {
+    reader.cancel().catch(() => {
+      return;
+    });
+    if (error instanceof HttpError)
+      throw error;
+    throw new HttpError("network", error instanceof Error ? error.message : String(error), timeoutMs, error);
+  } finally {
+    if (signal !== undefined && onAbort !== null)
+      signal.removeEventListener("abort", onAbort);
+  }
+}
+async function fetchTextWithTimeout(fetchImpl, url, init, options) {
+  const { timeoutMs } = options;
+  const controller = new AbortController;
+  const timer = setTimeout(() => controller.abort(), timeoutMs);
+  try {
+    let response;
+    try {
+      response = await fetchImpl(url, { ...init, signal: controller.signal });
+    } catch (error) {
+      if (controller.signal.aborted) {
+        throw new HttpError("timeout", `timed out after ${timeoutMs}ms: ${url}`, timeoutMs);
+      }
+      throw new HttpError("network", error instanceof Error ? error.message : String(error), timeoutMs, error);
+    }
+    const text = await readBodyText(response, {
+      signal: controller.signal,
+      timeoutMs,
+      ...options.maxBytes === undefined ? {} : { maxBytes: options.maxBytes }
+    });
+    return { response, text };
+  } finally {
+    clearTimeout(timer);
+  }
+}
+function defaultFetch() {
+  if (typeof globalThis.fetch !== "function") {
+    throw new TypeError("no global fetch available — pass one via createErc8004Client({fetch})");
+  }
+  return (input, init) => globalThis.fetch(input, init);
+}
+// ../../packages/graphql/dist/json.js
+function isRecord(value) {
+  return typeof value === "object" && value !== null && !Array.isArray(value);
+}
+function readRecord(source, key) {
+  if (!isRecord(source))
+    return null;
+  const child = source[key];
+  return isRecord(child) ? child : null;
+}
+function readString(source, key) {
+  if (!isRecord(source))
+    return null;
+  const value = source[key];
+  return typeof value === "string" ? value : null;
+}
+function readArray(source, key) {
+  if (!isRecord(source))
+    return [];
+  const value = source[key];
+  return Array.isArray(value) ? value : [];
+}
+// ../../packages/graphql/dist/client.js
+class GraphqlError extends Error {
+  url;
+  httpStatus;
+  constructor(message, url, httpStatus, cause) {
+    super(message, cause === undefined ? undefined : { cause });
+    this.name = "GraphqlError";
+    this.url = url;
+    this.httpStatus = httpStatus;
+  }
+}
+function createGraphqlTransport(config) {
+  const timeoutMs = config.timeoutMs ?? DEFAULT_TIMEOUT_MS;
+  return {
+    async request(query, variables, options) {
+      let response;
+      let text;
+      try {
+        ({ response, text } = await fetchTextWithTimeout(config.fetch, config.url, {
+          method: "POST",
+          headers: {
+            "content-type": "application/json",
+            accept: "application/json"
+          },
+          body: JSON.stringify({ query, variables })
+        }, {
+          timeoutMs: options?.timeoutMs ?? timeoutMs,
+          ...config.maxBodyBytes === undefined ? {} : { maxBytes: config.maxBodyBytes }
+        }));
+      } catch (error) {
+        const message = error instanceof HttpError ? error.message : String(error);
+        throw new GraphqlError(message, config.url, null, error);
+      }
+      if (!response.ok) {
+        throw new GraphqlError(`HTTP ${response.status} ${response.statusText}`, config.url, response.status);
+      }
+      let body = null;
+      try {
+        body = JSON.parse(text);
+      } catch {
+        body = null;
+      }
+      if (!isRecord(body)) {
+        throw new GraphqlError("response body is not a JSON object", config.url, 200);
+      }
+      const errors = readArray(body, "errors");
+      if (errors.length > 0) {
+        const messages = errors.map((entry) => readString(entry, "message") ?? "unknown error").join("; ");
+        throw new GraphqlError(messages, config.url, 200);
+      }
+      const data = body["data"];
+      if (!isRecord(data)) {
+        throw new GraphqlError("response carries no data object", config.url, 200);
+      }
+      return data;
+    }
+  };
+}
+// ../../packages/graphql/dist/queries/agent.js
+var THING = `value { thing { name description image url } }`;
+var MARKET_SIDE = `
+    total_market_cap
+    total_assets
+    vaults(where: {curve_id: {_eq: $curveId}}, limit: 1) {
+      market_cap
+      total_assets
+      total_shares
+      current_share_price
+      position_count
+    }`;
+var RESOLVE_AGENT_QUERY = `
+query ResolveAgentByCaip($sameAsPredicateId: String!, $caipId: String!) {
+  triples(
+    where: {
+      predicate_id: {_eq: $sameAsPredicateId}
+      object: {value: {thing: {name: {_eq: $caipId}}}}
+    }
+    limit: 2
+  ) {
+    subject { term_id label data ${THING} }
+  }
+}`;
+var TRUST_SURFACE_QUERY = `
+query AgentTrustSurface($subjectId: String!, $predicateIds: [String!], $curveId: numeric!, $limit: Int!) {
+  triples(
+    where: {subject_id: {_eq: $subjectId}, predicate_id: {_in: $predicateIds}}
+    order_by: {term: {total_market_cap: desc}}
+    limit: $limit
+  ) {
+    term_id
+    predicate { term_id label }
+    object { term_id label data ${THING} }
+    term {${MARKET_SIDE}
+    }
+    counter_term {${MARKET_SIDE}
+    }
+  }
+}`;
+var CAPABILITIES_QUERY = `
+query AgentCapabilities($subjectId: String!, $predicateIds: [String!], $limit: Int!) {
+  triples(
+    where: {subject_id: {_eq: $subjectId}, predicate_id: {_in: $predicateIds}}
+    order_by: {term_id: asc}
+    limit: $limit
+  ) {
+    term_id
+    predicate { term_id label }
+    object { term_id label ${THING} }
+  }
+}`;
+var IDENTITY_SAMPLE_LIMIT = 4;
+var CAPABILITY_SAMPLE_LIMIT = 8;
+var COHORT_QUERY = `
+query AgentCohort($where: triples_bool_exp!, $sameAsPredicateId: String!, $capabilityPredicateIds: [String!], $orderBy: [triples_order_by!]!, $limit: Int!, $offset: Int!) {
+  total: triples_aggregate(
+    where: $where
+  ) {
+    aggregate { count }
+  }
+  triples(
+    where: $where
+    order_by: $orderBy
+    limit: $limit
+    offset: $offset
+  ) {
+    subject {
+      term_id
+      label
+      ${THING}
+      as_subject_triples_aggregate { aggregate { count } }
+      term {
+        total_market_cap
+        vaults_aggregate { aggregate { count sum { position_count } } }
+      }
+      identityCount: as_subject_triples_aggregate(
+        where: {predicate_id: {_eq: $sameAsPredicateId}}
+      ) {
+        aggregate { count }
+      }
+      identity: as_subject_triples(
+        where: {predicate_id: {_eq: $sameAsPredicateId}}
+        limit: ${IDENTITY_SAMPLE_LIMIT}
+      ) {
+        object { label ${THING} }
+      }
+      capabilities: as_subject_triples(
+        where: {predicate_id: {_in: $capabilityPredicateIds}}
+        limit: ${CAPABILITY_SAMPLE_LIMIT}
+      ) {
+        predicate_id
+        object { term_id label }
+      }
+    }
+  }
+}`;
+// ../../packages/graphql/dist/queries/atom-standing.js
+var VAULT_SIDE = `
+    total_market_cap
+    total_assets
+    vaults(where: {curve_id: {_eq: $curveId}}, limit: 1) {
+      market_cap
+      total_assets
+      total_shares
+      current_share_price
+      position_count
+    }`;
+var ATOM_STANDING_QUERY = `
+query AtomStanding($atomId: String!, $curveId: numeric!, $limit: Int!, $attestor: [String!]!) {
+  atom: atoms(where: {term_id: {_eq: $atomId}}, limit: 1) {
+    term_id
+    label
+    data
+    type
+    value { thing { url } }
+    term { ${VAULT_SIDE} }
+  }
+  triples(
+    where: {object_id: {_eq: $atomId}}
+    order_by: {term: {total_assets: desc}}
+    limit: $limit
+  ) {
+    term_id
+    predicate { term_id label }
+    subject { term_id label type }
+    term { ${VAULT_SIDE} }
+    counter_term { ${VAULT_SIDE} }
+  }
+  claims: triples(
+    where: {subject_id: {_eq: $atomId}}
+    order_by: {term: {total_assets: desc}}
+    limit: $limit
+  ) {
+    term_id
+    predicate { term_id label }
+    object { term_id label type }
+    term {
+      ${VAULT_SIDE}
+      vaults(where: {curve_id: {_eq: $curveId}}, limit: 1) {
+        positions(where: {account_id: {_in: $attestor}, shares: {_gt: "0"}}, limit: 1) { shares }
+      }
+    }
+  }
+}`;
+var LISTED_CAPABILITIES_FOR_QUERY = `query ListedCapabilitiesFor($subject: String!, $predicates: [String!]) {
+  atoms(where: {term_id: {_eq: $subject}}, limit: 1) {
+    capabilities: as_subject_triples(
+      where: {predicate_id: {_in: $predicates}}
+      limit: ${CAPABILITY_SAMPLE_LIMIT}
+    ) { object { term_id } }
+  }
+}`;
+// ../../packages/graphql/dist/queries/tool-listing.js
+function toolListingQuery(sponsored) {
+  const variables = sponsored ? "$uses: String!, $sameAs: String!, $curveId: numeric!, $limit: Int!, $sponsor: String!" : "$uses: String!, $sameAs: String!, $curveId: numeric!, $limit: Int!";
+  const sponsoredCount = sponsored ? `
+      sponsored: deposits_aggregate(where: {sender_id: {_ilike: $sponsor}, receiver_id: {_nilike: $sponsor}}) { aggregate { count } }` : "";
+  return `
+query ToolListing(${variables}) {
+  triples(
+    where: {
+      predicate_id: {_eq: $uses}
+      subject: {as_subject_triples: {predicate_id: {_eq: $sameAs}}}
+    }
+    limit: $limit
+  ) {
+    term_id
+    subject_id
+    object {
+      term_id
+      label
+      data
+      type
+      value { thing { url } }
+    }
+    term {
+      total_assets
+      vaults(where: {curve_id: {_eq: $curveId}}, limit: 1) { position_count total_assets }
+      deposits: deposits_aggregate { aggregate { count } }${sponsoredCount}
+    }
+    counter_term {
+      total_assets
+      vaults(where: {curve_id: {_eq: $curveId}}, limit: 1) { position_count total_assets }
+    }
+  }
+}`;
+}
+var TOOL_LISTING_QUERY = toolListingQuery(false);
+var TOOL_LISTING_SPONSORED_QUERY = toolListingQuery(true);
+// ../../packages/graphql/dist/queries/pulse-usage.js
+var PULSE_USES_QUERY = `query PulseUses($tools: [String!]!, $uses: String!, $limit: Int!) {
+    triples(where: {object_id: {_in: $tools}, predicate_id: {_eq: $uses}}, limit: $limit) {
+        term_id
+        object_id
+        creator_id
+        created_at
+        subject {
+            term_id
+            label
+        }
+        positions(where: {shares: {_gt: "0"}}) {
+            account_id
+        }
+    }
+}`;
+var PULSE_QUALIFIERS_QUERY = `query PulseQualifiers($edges: [String!]!, $predicates: [String!]!, $writers: [String!]!, $limit: Int!) {
+    triples(where: {subject_id: {_in: $edges}, predicate_id: {_in: $predicates}}, limit: $limit) {
+        term_id
+        subject_id
+        predicate_id
+        creator_id
+        created_at
+        object {
+            label
+        }
+        positions(where: {shares: {_gt: "0"}, account_id: {_in: $writers}}) {
+            account_id
+        }
+    }
+}`;
+var PULSE_ATOMS_QUERY = `query PulseAtoms($uris: [String!]!, $limit: Int!) {
+    atoms(where: {data: {_in: $uris}}, limit: $limit) {
+        term_id
+        data
+    }
+}`;
+// ../../packages/graphql/dist/queries/tool-usage.js
+var VAULT_SIDE2 = `
+    total_market_cap
+    total_assets
+    vaults(where: {curve_id: {_eq: $curveId}}, limit: 1) {
+      market_cap
+      total_assets
+      total_shares
+      current_share_price
+      position_count
+    }`;
+var TOOL_USAGE_QUERY = `
+query ToolUsage($toolId: String!, $uses: String!, $curveId: numeric!) {
+  tool: atoms(where: {term_id: {_eq: $toolId}}, limit: 1) {
+    term_id
+    label
+    data
+  }
+  triples(where: {predicate_id: {_eq: $uses}, object_id: {_eq: $toolId}}, limit: 200) {
+    term_id
+    subject {
+      term_id
+      label
+      type
+    }
+    term { ${VAULT_SIDE2} }
+    counter_term { ${VAULT_SIDE2} }
+  }
+}`;
+var ATTESTED_TOOL_USAGE_QUERY = `
+query AttestedToolUsage($toolId: String!, $uses: String!, $curveId: numeric!, $attestor: String!) {
+  tool: atoms(where: {term_id: {_eq: $toolId}}, limit: 1) {
+    term_id
+    label
+    data
+  }
+  triples(where: {
+    predicate_id: {_eq: $uses},
+    object_id: {_eq: $toolId},
+    term: {positions: {account_id: {_eq: $attestor}}}
+  }, limit: 200) {
+    term_id
+    subject {
+      term_id
+      label
+      type
+    }
+    term { ${VAULT_SIDE2} }
+    counter_term { ${VAULT_SIDE2} }
+  }
+}`;
+// ../../packages/graphql/dist/queries/x402-feed.js
+var LEVEL = `term_id
+        subject_id
+        predicate_id
+        creator_id
+        created_at
+        subject {
+            term_id
+            label
+        }
+        object {
+            term_id
+            label
+            data
+            value {
+                thing {
+                    url
+                }
+            }
+        }
+        positions(where: {shares: {_gt: "0"}}) {
+            account_id
+        }`;
+var X402_PAYMENT_LEVELS_QUERY = `query X402PaymentLevels($paidWith: String!, $limit: Int!, $offset: Int!) {
+    triples(where: {predicate_id: {_eq: $paidWith}}, order_by: {term_id: asc}, limit: $limit, offset: $offset) {
+        ${LEVEL}
+    }
+}`;
+var X402_LEVELS_BY_ID_QUERY = `query X402LevelsById($ids: [String!]!, $limit: Int!) {
+    triples(where: {term_id: {_in: $ids}}, limit: $limit) {
+        ${LEVEL}
+    }
+}`;
+var X402_REASONS_QUERY = `query X402Reasons($ids: [String!]!, $because: String!, $limit: Int!) {
+    triples(where: {subject_id: {_in: $ids}, predicate_id: {_eq: $because}}, order_by: {term_id: asc}, limit: $limit) {
+        ${LEVEL}
+    }
+}`;
 // ../../packages/protocol/dist/declare/tool-thing.js
 class InvalidToolUrlError extends Error {
   input;
@@ -211,140 +698,6 @@ function readRequires(markdown) {
   }
   return { cli, runs, login: null, description };
 }
-// ../../packages/protocol/dist/skill-domains.js
-var SKILL_DOMAINS = [
-  {
-    domain: "payment",
-    parent: "tool",
-    words: ["payment", "stripe", "paypal", "invoice", "checkout", "billing", "paywall"]
-  },
-  {
-    domain: "video",
-    parent: "tool",
-    words: ["video", "remotion", "hyperframes", "keyframe", "footage", "youtube"],
-    nameWords: ["render", "caption", "motion", "title", "carousel", "cursor", "frame", "overlay", "shard", "animation", "avatar", "clip"]
-  },
-  {
-    domain: "image",
-    parent: "tool",
-    words: ["image generation", "inpainting", "outpainting", "flux", "midjourney", "stable diffusion", "nano banana"],
-    nameWords: ["image", "photo", "relight", "upscale", "thumbnail", "imagegen"]
-  },
-  {
-    domain: "audio",
-    parent: "tool",
-    words: ["elevenlabs", "music generation", "text to speech", "voiceover", "sound effect", "instrumental"],
-    nameWords: ["audio", "music", "voice", "speech", "sound"]
-  },
-  {
-    domain: "document",
-    parent: "tool",
-    words: ["pptx", "docx", "xlsx", "powerpoint", "spreadsheet", ".potx", "pdf file"],
-    nameWords: ["pdf", "docx", "pptx", "xlsx", "document", "slide", "markdown"]
-  },
-  {
-    domain: "scraping",
-    parent: "tool",
-    words: ["scrap", "crawl", "firecrawl", "apify"],
-    nameWords: ["extract", "scrape"]
-  },
-  {
-    domain: "browser",
-    parent: "tool",
-    words: ["playwright", "puppeteer", "headless browser", "selenium"],
-    nameWords: ["browser", "chrome"]
-  },
-  {
-    domain: "database",
-    parent: "tool",
-    words: ["database", "postgres", "prisma", "supabase", "mongo", " sql"],
-    nameWords: ["schema", "query", "migration", "sql"]
-  },
-  {
-    domain: "design",
-    parent: "tool",
-    words: ["figma", "typograph", "design system", "design token"],
-    nameWords: ["design", "ux", "colour", "color", "layout", "logo", "icon", "theme"]
-  },
-  {
-    domain: "collaboration",
-    parent: "tool",
-    words: ["slack", "notion", "jira", "linear", "atlassian", "monday", "飞书", "obsidian"],
-    nameWords: ["lark", "calendar", "whiteboard", "meeting"]
-  },
-  {
-    domain: "marketing",
-    parent: "tool",
-    words: ["marketing", "seo", "campaign", "copywriting", "newsletter", "onboarding", "referral", "product hunt"],
-    nameWords: ["cro", "launch", "photography", "signup", "popup", "ads", "audience", "growth", "brand"]
-  },
-  {
-    domain: "blockchain",
-    parent: "tool",
-    words: ["blockchain", "onchain", "on-chain", "ethereum", "solidity", "web3", "smart contract", "erc-20", "erc-721"],
-    nameWords: ["wallet", "crypto", "nft", "defi"]
-  },
-  {
-    domain: "infrastructure",
-    parent: "tool",
-    words: ["deploy", "docker", "kubernetes", "terraform", "cloudflare", "aws", "azure", "serverless", "ci/cd", "firebase", "hosting", "vercel"],
-    nameWords: ["infra", "cli"]
-  },
-  {
-    domain: "observability",
-    parent: "tool",
-    words: ["observab", "logging", "grafana", "sentry", "telemetry"],
-    nameWords: ["monitor", "metric", "trace", "alert", "log"]
-  },
-  {
-    domain: "quality",
-    parent: "tool",
-    words: ["linting", "unit test", "code coverage", "static analysis"],
-    nameWords: ["test", "lint", "review", "debug", "audit", "coverage", "quality", "refactor", "eval"]
-  },
-  {
-    domain: "writing",
-    parent: "tool",
-    words: ["cold email", "pr body", "commit message", "conventional commits"],
-    nameWords: ["writing", "write", "article", "prose", "caveman", "cave", "edit"]
-  },
-  {
-    domain: "method",
-    parent: "tool",
-    words: ["handoff", "hand the current", "implementation plan", "subagent", "superpowers", "overbuilding", "chief of staff", "retrospective", "acceptance condition", "router over", "re-pitch"],
-    nameWords: ["handoff", "triage", "retro", "brainstorm", "spec", "prd", "questionnaire", "plan", "implement", "dispatch", "execute", "executing", "investigate", "wayfinder", "decision", "teach", "verify", "verification", "authoring", "template", "zoom", "wait", "grill", "ask", "manage", "optimize", "worktree", "lean", "sensei", "help", "ticket", "issue"]
-  },
-  {
-    domain: "website",
-    parent: "tool",
-    words: ["landing page", "static site", "next.js", "nextjs", "tailwind"],
-    nameWords: ["website", "html", "css", "landing", "frontend", "react"]
-  },
-  {
-    domain: "api",
-    parent: "tool",
-    words: ["rest api", "graphql", "webhook", "openapi", "postman"],
-    nameWords: ["api", "endpoint", "rest"]
-  },
-  {
-    domain: "automation",
-    parent: "tool",
-    words: ["zapier", "cron", "orchestrat"],
-    nameWords: ["automation", "workflow", "schedule", "pipeline"]
-  },
-  {
-    domain: "development",
-    parent: "tool",
-    words: ["typescript", "python", "rust ", "sdk"],
-    nameWords: ["code", "compile", "library", "framework", "git"]
-  }
-];
-function classifySkill(skill, description) {
-  const name = ` ${skill.replace(/-/g, " ").toLowerCase()} `;
-  const both = `${name}${description.toLowerCase()}`;
-  const hit = SKILL_DOMAINS.find((rule) => rule.words.some((word) => both.includes(word)) || (rule.nameWords ?? []).some((word) => name.includes(` ${word}`)));
-  return hit?.domain ?? null;
-}
 // ../../packages/protocol/dist/content-path.js
 function skillRefFromUrl(url) {
   let parsed;
@@ -395,12 +748,6 @@ function installedFromReceipts(text) {
   }
   return found;
 }
-// ../../packages/protocol/dist/pulse-score.js
-function inferDomain(query, labels) {
-  const words = ` ${query.toLowerCase().replace(/[^a-z0-9-]+/g, " ").trim()} `;
-  const matches = labels.filter((label) => words.includes(` ${label.toLowerCase()} `));
-  return [...matches].sort((a, b) => b.length - a.length)[0] ?? null;
-}
 // ../hook/src/classify.ts
 var CANCELLED = /\b(interrupted|cancel(l)?ed|aborted) by (the )?user\b|\buser (rejected|denied|declined)\b|\bpermission denied by user\b|\buser doesn't want to (proceed|take this action)\b/i;
 var AUTH = /\b(unauthori[sz]ed|forbidden|invalid (api )?(key|token)|authentication|not authenticated|expired token)\b/i;
@@ -408,7 +755,7 @@ var TIMEOUT = /\b(timed? ?out|timeout|deadline exceeded|ETIMEDOUT)\b/i;
 var UNREACHABLE = /\b(ENOTFOUND|EAI_AGAIN|ECONNREFUSED|ECONNRESET|EHOSTUNREACH|ENETUNREACH|getaddrinfo|certificate|TLS|SSL|socket hang up|connection refused|fetch failed)\b/i;
 var PAYMENT = /\b(payment required|insufficient (funds|balance)|payment (failed|refused|rejected)|x402)\b/i;
 var STATUS = /\b(40[123]|429|5\d\d)\b/;
-function isRecord(value) {
+function isRecord2(value) {
   return typeof value === "object" && value !== null && !Array.isArray(value);
 }
 function failureText(call) {
@@ -416,7 +763,7 @@ function failureText(call) {
   const push = (value) => {
     if (typeof value === "string")
       parts.push(value);
-    else if (isRecord(value)) {
+    else if (isRecord2(value)) {
       for (const key of ["error", "message", "reason"]) {
         const text = value[key];
         if (typeof text === "string")
@@ -424,7 +771,7 @@ function failureText(call) {
       }
       if (Array.isArray(value.content)) {
         for (const item of value.content) {
-          if (isRecord(item) && typeof item.text === "string")
+          if (isRecord2(item) && typeof item.text === "string")
             parts.push(item.text);
         }
       }
@@ -436,7 +783,7 @@ function failureText(call) {
 `).slice(0, 4000);
 }
 function statusOf(call) {
-  if (!isRecord(call.response))
+  if (!isRecord2(call.response))
     return null;
   for (const key of ["status", "statusCode", "status_code"]) {
     const value = call.response[key];
@@ -446,7 +793,7 @@ function statusOf(call) {
   return null;
 }
 function flaggedError(response) {
-  return isRecord(response) && (response.isError === true || response.is_error === true);
+  return isRecord2(response) && (response.isError === true || response.is_error === true);
 }
 function classifyCall(call) {
   if (call.interrupted)
@@ -476,7 +823,7 @@ function classifyCall(call) {
 import { readFileSync } from "node:fs";
 import { homedir } from "node:os";
 import { join } from "node:path";
-function isRecord2(value) {
+function isRecord3(value) {
   return typeof value === "object" && value !== null && !Array.isArray(value);
 }
 function readJson(path) {
@@ -487,7 +834,7 @@ function readJson(path) {
   }
 }
 function entryOf(raw) {
-  if (!isRecord2(raw))
+  if (!isRecord3(raw))
     return null;
   if (typeof raw.url === "string" && raw.url.length > 0)
     return { transport: "remote", url: raw.url };
@@ -498,7 +845,7 @@ function entryOf(raw) {
   return null;
 }
 function collect(into, servers) {
-  if (!isRecord2(servers))
+  if (!isRecord3(servers))
     return;
   for (const [name, raw] of Object.entries(servers)) {
     const entry = entryOf(raw);
@@ -512,16 +859,16 @@ function readMcpServers({
 }) {
   const servers = new Map;
   const user = readJson(join(home, ".claude.json"));
-  if (isRecord2(user)) {
+  if (isRecord3(user)) {
     collect(servers, user.mcpServers);
-    if (isRecord2(user.projects)) {
+    if (isRecord3(user.projects)) {
       const project = user.projects[cwd];
-      if (isRecord2(project))
+      if (isRecord3(project))
         collect(servers, project.mcpServers);
     }
   }
   const shared = readJson(join(cwd, ".mcp.json"));
-  if (isRecord2(shared))
+  if (isRecord3(shared))
     collect(servers, shared.mcpServers);
   return servers;
 }
@@ -644,7 +991,7 @@ function capture(payload, dir, now = new Date) {
 }
 
 // ../hook/src/payload.ts
-function isRecord3(value) {
+function isRecord4(value) {
   return typeof value === "object" && value !== null && !Array.isArray(value);
 }
 function str(value) {
@@ -659,8 +1006,13 @@ function first(raw, keys) {
       return raw[key];
   return;
 }
+function eventOf(raw) {
+  if (!isRecord4(raw))
+    return null;
+  return str(first(raw, ["hook_event_name", "hookEventName"]));
+}
 function toolCallOf(raw) {
-  if (!isRecord3(raw))
+  if (!isRecord4(raw))
     return null;
   const format = typeof raw.toolName === "string" && typeof raw.tool_name !== "string" ? "grok" : "claude";
   const camel = format === "grok";
@@ -676,7 +1028,7 @@ function toolCallOf(raw) {
     session: str(camel ? raw.sessionId : raw.session_id),
     id: str(camel ? raw.toolUseId : raw.tool_use_id),
     tool,
-    input: isRecord3(input) ? input : undefined,
+    input: isRecord4(input) ? input : undefined,
     hasResponse,
     response: hasResponse ? first(raw, responseKeys) : undefined,
     error: str(raw.error),
@@ -687,60 +1039,2788 @@ function toolCallOf(raw) {
   };
 }
 
-// ../hook/src/skill-failure.ts
-import { existsSync, mkdirSync as mkdirSync2, readFileSync as readFileSync2 } from "node:fs";
-import { homedir as homedir2 } from "node:os";
-import { dirname as dirname2, join as join3 } from "node:path";
-var MISSING_KEY = /\b(api[ _-]?key|access[ _-]?token|credentials?)\b.{0,40}\b(missing|not set|not found|required|invalid|expired)\b|(?<!\bnot )\b(missing|no|invalid|expired)\b.{0,20}\b(api[ _-]?key|access[ _-]?token|credentials?)\b|\b(HTTP|status)(\/[\d.]+| code)?[ :=]*(401|403)\b|\b(401|403) (unauthori[sz]ed|forbidden)\b|\bunauthori[sz]ed\b|not authenticated|not logged in/i;
-var MISSING_KEY_VARIABLE = /\b[A-Z0-9_]*(API_KEY|TOKEN|SECRET)\b.{0,30}\b(not set|missing|required|unset)\b/;
-var MISSING_PROGRAM = /command not found|executable file not found|\bnot installed\b|No module named|spawn \S+ ENOENT|Cannot find module/i;
-var MISSING_FILE = /No such file or directory|\bENOENT\b|file not found|\b(file|path|directory|folder)\b(?!\.).{0,80}\bdoes not exist\b/i;
-function failureReason(text) {
-  if (MISSING_PROGRAM.test(text))
-    return "missing-program";
-  if (MISSING_KEY.test(text) || MISSING_KEY_VARIABLE.test(text))
-    return "missing-key";
-  if (MISSING_FILE.test(text))
-    return "missing-file";
-  return null;
+// ../../packages/protocol/src/content-path.ts
+function skillRefFromUrl3(url) {
+  let parsed;
+  try {
+    parsed = new URL(url.includes("://") ? url : `https://${url}`);
+  } catch {
+    return null;
+  }
+  if (parsed.hostname.replace(/^www\./, "") !== "github.com")
+    return null;
+  const skill = parsed.searchParams.get("skill");
+  if (skill === null || skill.length === 0)
+    return null;
+  const [owner, repo] = parsed.pathname.replace(/^\/+/, "").replace(/\/+$/, "").replace(/\.git$/, "").split("/");
+  if (owner === undefined || repo === undefined)
+    return null;
+  if (owner.length === 0 || repo.length === 0)
+    return null;
+  return { repo: `${owner}/${repo}`, skill };
 }
-var RUNNERS = new Set(["bash", "bun", "bunx", "node", "npx", "python", "python3", "sh", "uv", "uvx", "zsh"]);
-function programsOf(command) {
+function directoryOf2(path) {
+  const parts = path.split("/");
+  return parts.length < 2 ? "" : parts[parts.length - 2] ?? "";
+}
+function skillFilesIn2(listing) {
+  return listing.filter((file) => (file.path.split("/").pop() ?? "").toLowerCase() === "skill.md");
+}
+
+// ../../packages/protocol/src/skill-requires.ts
+var LOGIN2 = /\b(auth login|log ?in|sign ?up|api[ _-]?key|access token)\b/i;
+function frontmatter2(markdown) {
+  const lines = markdown.split(`
+`);
+  if (lines[0]?.trim() !== "---")
+    return { lines: [], bodyStart: 0 };
+  const end = lines.findIndex((line, index) => index > 0 && line.trim() === "---");
+  return end === -1 ? { lines: [], bodyStart: 0 } : { lines: lines.slice(1, end), bodyStart: end + 1 };
+}
+function unquote2(value) {
+  return value.trim().replace(/^["']|["']$/g, "");
+}
+function declaredBins2(lines) {
+  const at = lines.findIndex((line) => /^\s*bins\s*:/.test(line));
+  if (at === -1)
+    return [];
+  const inline = lines[at]?.split(":").slice(1).join(":").trim() ?? "";
+  if (inline.startsWith("[")) {
+    return inline.replace(/^\[|\]$/g, "").split(",").map(unquote2).filter((name) => name.length > 0);
+  }
+  const out = [];
+  for (const line of lines.slice(at + 1)) {
+    const item = /^\s*-\s*(.+)$/.exec(line);
+    if (item === null)
+      break;
+    out.push(unquote2(item[1] ?? ""));
+  }
+  return out.filter((name) => name.length > 0);
+}
+function bashTools2(lines) {
+  const line = lines.find((candidate) => /^\s*allowed-tools\s*:/.test(candidate));
+  if (line === undefined)
+    return [];
+  return [...line.matchAll(/Bash\(\s*([A-Za-z0-9._-]+)/g)].map((match) => match[1] ?? "").filter((name) => name.length > 0);
+}
+function declaredDescription2(lines) {
+  const at = lines.findIndex((line) => /^description\s*:/.test(line));
+  if (at === -1)
+    return null;
+  const inline = lines[at]?.split(":").slice(1).join(":").trim() ?? "";
+  if (inline !== "" && !/^[>|][-+]?$/.test(inline))
+    return unquote2(inline) || null;
+  const block = [];
+  for (const line of lines.slice(at + 1)) {
+    if (!/^\s+\S/.test(line))
+      break;
+    block.push(line.trim());
+  }
+  return block.length === 0 ? null : block.join(" ");
+}
+var COMMON2 = new Set(("awk bash break bun case cat cd chmod command continue cp curl cut date diff do done echo else end env " + "esac exit export false fi find for function git grep head if kill ln local ls mkdir mv node npm npx open pip pip3 pnpm printf pwd " + "python python3 read return rm sed set sh shift sleep sort source sudo tail tar tee test then time " + "touch tr true uniq unset unzip wait wc wget which while xargs yarn zip").split(" "));
+var INSTALL2 = /\b(?:npm\s+(?:i|install)\s+(?:-g|--global)|pnpm\s+add\s+-g|yarn\s+global\s+add|brew\s+install|pipx\s+install|pip3?\s+install|cargo\s+install)\s+([@a-z0-9][@\w./-]*)/i;
+var PROBE2 = /\b(?:command\s+-v|which)\s+([a-z][\w-]*)/i;
+var COMMAND2 = /^[a-z][a-z0-9_-]*$/;
+function commandsIn2(shell, explicitOnly) {
   const found = [];
-  for (const part of command.split(/&&|\|\||[;|\n]/)) {
-    const words = part.trim().split(/\s+/).filter((word) => !/^[A-Za-z_][A-Za-z0-9_]*=/.test(word));
-    const first2 = words[0];
-    if (first2 === undefined || first2 === "")
+  for (const segment of shell.split(/&&|\|\||[|;]/)) {
+    const line = segment.trim().replace(/^\$\s+/, "");
+    if (line === "" || line.startsWith("#"))
       continue;
-    const program = first2.split("/").pop() ?? first2;
-    if (!RUNNERS.has(program)) {
-      found.push(program);
+    const named = INSTALL2.exec(line)?.[1] ?? PROBE2.exec(line)?.[1];
+    if (named !== undefined) {
+      found.push(named);
       continue;
     }
-    const target = words.slice(1).find((word) => !word.startsWith("-") && word !== "run");
-    if (target !== undefined)
-      found.push(target.split("/").pop() ?? target);
+    if (explicitOnly)
+      continue;
+    const words = line.split(/\s+/).filter((word) => !/^[A-Za-z_][A-Za-z0-9_]*=/.test(word));
+    const first2 = words[0] === "sudo" ? words[1] : words[0];
+    if (first2 === undefined || !COMMAND2.test(first2))
+      continue;
+    found.push(first2);
   }
   return found;
 }
-var escaped = (text) => text.replace(/[.*+?^${}()|[\]\\]/g, "\\$&");
-function askedBySkill(command, markdown, directory) {
-  if (new RegExp(`${escaped(directory)}(?=[/\\s'"]|$)`).test(command))
-    return true;
-  const { cli, runs } = readRequires(markdown);
-  const named = new Set([...cli, ...runs]);
-  return programsOf(command).some((program) => named.has(program) || program.includes(".") && new RegExp(`(?:^|[\\s\`'"/(])${escaped(program)}(?=[\\s\`'")]|$)`, "m").test(markdown));
-}
-var DOMAIN = /^[a-z0-9][a-z0-9-]*$/;
-var DOMAINS = [...new Set(SKILL_DOMAINS.map((rule) => rule.domain))];
-function lastUserWords(transcript) {
-  let words = "";
-  for (const line of transcript.split(`
-`)) {
-    if (!line.includes('"user"'))
+var SHELL_FENCE2 = /^(sh|bash|shell|zsh|console|terminal)$/i;
+function shellLinesRun2(lines) {
+  const run = [];
+  let heredocEnd = null;
+  let quote = null;
+  let continued = false;
+  for (const line of lines) {
+    if (heredocEnd !== null) {
+      if (line.trim() === heredocEnd)
+        heredocEnd = null;
       continue;
-    try {
-      const entry = JSON.parse(line);
+    }
+    if (quote === null && !continued)
+      run.push(line);
+    continued = line.trimEnd().endsWith("\\");
+    heredocEnd = /<<-?\s*['"]?([A-Za-z_]+)/.exec(line)?.[1] ?? null;
+    for (const char of line) {
+      if (quote === null && (char === '"' || char === "'"))
+        quote = char;
+      else if (char === quote)
+        quote = null;
+    }
+  }
+  return run;
+}
+function commandsRun2(body, declared) {
+  const found = [];
+  let fence = null;
+  let fenced = [];
+  for (const line of body) {
+    const marker = /^\s*(?:```|~~~)\s*([\w-]*)/.exec(line);
+    if (marker !== null) {
+      if (fence === "shell") {
+        for (const run of shellLinesRun2(fenced))
+          found.push(...commandsIn2(run, false));
+      }
+      fenced = [];
+      fence = fence !== null ? null : SHELL_FENCE2.test(marker[1] ?? "") ? "shell" : "other";
+      continue;
+    }
+    if (fence !== null) {
+      fenced.push(line);
+      continue;
+    }
+    for (const inline of line.matchAll(/`([^`]+)`/g))
+      found.push(...commandsIn2(inline[1] ?? "", true));
+  }
+  return [...new Set(found)].filter((name) => !COMMON2.has(name) && !declared.includes(name));
+}
+function requiresLine2(lines) {
+  const name = lines.map((line) => /^\s*requires\s*:\s*(\S+)\s*$/.exec(line)?.[1]).find(Boolean);
+  return name === undefined ? [] : [unquote2(name)];
+}
+function readRequires2(markdown) {
+  const { lines, bodyStart } = frontmatter2(markdown);
+  const cli = [...new Set([...declaredBins2(lines), ...bashTools2(lines), ...requiresLine2(lines)])];
+  const description = declaredDescription2(lines);
+  const body = markdown.split(`
+`);
+  const runs = commandsRun2(body.slice(bodyStart), cli);
+  for (let index = bodyStart;index < body.length; index += 1) {
+    const text = body[index] ?? "";
+    if (LOGIN2.test(text)) {
+      return { cli, runs, login: { line: index + 1, text: text.trim().slice(0, 160) }, description };
+    }
+  }
+  return { cli, runs, login: null, description };
+}
+
+// ../../packages/protocol/src/skill-contents.ts
+var CONTENTS_LIMITS2 = {
+  files: 200,
+  fileChars: 512 * 1024,
+  items: 40,
+  lineChars: 160
+};
+var SCRIPT = /\.(sh|bash|zsh|py|js|mjs|cjs|ts|rb|pl|ps1|php|lua)$/i;
+var MARKDOWN = /\.(md|mdx|markdown)$/i;
+var FENCE = /^\s*(?:```|~~~)/;
+function cap(text) {
+  const flat = text.replace(/\s+/g, " ").trim();
+  return flat.length > CONTENTS_LIMITS2.lineChars ? `${flat.slice(0, CONTENTS_LIMITS2.lineChars - 1)}…` : flat;
+}
+function compare(a, b) {
+  return a < b ? -1 : a > b ? 1 : 0;
+}
+var BASE_SHELLS = new Set(["sh", "bash", "zsh", "dash"]);
+function interpreterOf(firstLine) {
+  const match = /^#!\s*(\S+)(?:\s+(?:-\S+\s+)*(\S+))?/.exec(firstLine);
+  if (match === null)
+    return null;
+  const binary = match[1]?.split("/").pop() ?? "";
+  const name = binary === "env" ? match[2] ?? "" : binary;
+  return /^[a-z][\w.-]*$/i.test(name) && !BASE_SHELLS.has(name) ? name : null;
+}
+var SCRIPT_PROBE = /(?:^|[;&|(!{]|\bif|\bthen|\$\()\s*(?:!\s*)?(?:command\s+-v|which|type\s+-p|hash)\s+([a-z][\w.+-]*)/g;
+var SYSTEM_INSTALL = /\b(?:brew|apt-get|apt|dnf|yum|apk|pacman|choco|winget|scoop|port)\s+(?:install|add|-S)\s+(?:-\S+\s+)*([a-z][\w.+-]*)/g;
+var COMMENT_LINE = /^\s*(?:#|\/\/|\*|--|;|rem\s)/i;
+var PYTHON_PROBE = /\bshutil\.which\(\s*["']([a-z][\w.+-]*)["']/g;
+function programsInScriptLine(line, shell) {
+  if (COMMENT_LINE.test(line))
+    return [];
+  const matches = shell ? [...line.matchAll(SCRIPT_PROBE), ...line.matchAll(SYSTEM_INSTALL)] : [...line.matchAll(PYTHON_PROBE)];
+  return matches.map((match) => match[1] ?? "").filter((name) => name !== "");
+}
+var PACKAGE_INSTALL = /\b(pip3?|pipx|uv\s+pip|uv\s+add|poetry\s+add|npm|pnpm|yarn|bun|gem|cargo)\s+(?:install|add|i)\b([^\n|;&#`)]*)/gi;
+function packagesIn(line) {
+  const out = [];
+  for (const match of line.matchAll(PACKAGE_INSTALL)) {
+    const manager = (match[1] ?? "").replace(/\s+/g, " ").toLowerCase();
+    for (const raw of (match[2] ?? "").trim().split(/\s+/)) {
+      const word = raw.replace(/^["']|["']$/g, "");
+      if (word === "" || word.startsWith("-") || word.startsWith(".") || /[:\\<>{}$]/.test(word))
+        continue;
+      if (/\.(txt|toml|json|cfg|lock|whl|tgz|tar\.gz|zip)$/i.test(word))
+        continue;
+      const name = word.replace(/[=<>!~[].*$/, "");
+      if (/^(@[a-z0-9][\w.-]*\/)?[a-z0-9][\w.-]*(@[\w.^~-]+)?$/i.test(name))
+        out.push({ manager, name });
+    }
+  }
+  return out;
+}
+function requirementsIn(text) {
+  return text.split(`
+`).map((line) => line.replace(/#.*/, "").trim()).filter((line) => line !== "" && !line.startsWith("-") && !/^(https?|git\+)/.test(line)).map((line) => /^[A-Za-z0-9][\w.-]*/.exec(line)?.[0] ?? "").filter((name) => name !== "");
+}
+var SHELL_VARIABLES = new Set(("HOME PATH PWD OLDPWD USER LOGNAME SHELL TERM LANG LC_ALL TMPDIR TMP TEMP HOSTNAME IFS PS1 PS2 " + "RANDOM SECONDS LINENO BASH_SOURCE BASH_VERSION FUNCNAME PIPESTATUS OPTARG OPTIND EUID UID PPID " + "REPLY EDITOR VISUAL PAGER DISPLAY XDG_CONFIG_HOME XDG_DATA_HOME XDG_CACHE_HOME CI NODE_ENV DEBUG").split(" "));
+var READS_IN_CODE = [
+  /\bprocess\.env\.([A-Z][A-Z0-9_]+)/g,
+  /\bprocess\.env\[\s*["']([A-Z][A-Z0-9_]+)["']\s*\]/g,
+  /\bBun\.env\.([A-Z][A-Z0-9_]+)/g,
+  /\bDeno\.env\.get\(\s*["']([A-Z][A-Z0-9_]+)["']/g,
+  /\bimport\.meta\.env\.([A-Z][A-Z0-9_]+)/g,
+  /\bos\.environ\[\s*["']([A-Z][A-Z0-9_]+)["']\s*\]/g,
+  /\b(?:os\.environ\.get|os\.getenv|getenv|environ\.get)\(\s*["']([A-Z][A-Z0-9_]+)["']/g,
+  /\bENV(?:\.fetch\(|\[)\s*["']([A-Z][A-Z0-9_]+)["']/g,
+  /\$env:([A-Za-z][A-Za-z0-9_]+)/g
+];
+var SHELL_READ = /\$\{?([A-Z][A-Z0-9_]{2,})\b/g;
+var SHELL_ASSIGN = /^\s*(?:export\s+|local\s+|readonly\s+|declare\s+(?:-\w+\s+)?)?([A-Z][A-Z0-9_]*)=/;
+var SHELL_LOOP = /\b(?:for|read(?:\s+-\w+)*)\s+([A-Z][A-Z0-9_]*)\b/g;
+var EXPORT_IN_PROSE = /\bexport\s+([A-Z][A-Z0-9_]{2,})=/g;
+var KEY_NAMED = /\b([A-Z][A-Z0-9]*(?:_[A-Z0-9]+)*_(?:API_KEY|APIKEY|KEY|TOKEN|SECRET|PASSWORD))\b/g;
+function shellReads(lines) {
+  const assigned = new Set;
+  for (const line of lines) {
+    const name = SHELL_ASSIGN.exec(line)?.[1];
+    if (name !== undefined)
+      assigned.add(name);
+    for (const loop of line.matchAll(SHELL_LOOP))
+      assigned.add(loop[1] ?? "");
+  }
+  return lines.flatMap((line) => [...line.matchAll(SHELL_READ)].map((match) => match[1] ?? "")).filter((name) => !assigned.has(name));
+}
+function isShell(path, text) {
+  return /\.(sh|bash|zsh)$/i.test(path) || /^#!\s*\S*\/(?:env\s+)?(?:ba|z|da)?sh\b/.test(text);
+}
+function settingsIn(path, text, code) {
+  const found = new Set;
+  for (const pattern of READS_IN_CODE) {
+    for (const match of text.matchAll(pattern))
+      found.add((match[1] ?? "").toUpperCase());
+  }
+  if (MARKDOWN.test(path)) {
+    for (const match of text.matchAll(EXPORT_IN_PROSE))
+      found.add(match[1] ?? "");
+    for (const match of text.matchAll(KEY_NAMED))
+      found.add(match[1] ?? "");
+  } else if (isShell(path, text)) {
+    for (const name of shellReads(code))
+      found.add(name);
+  }
+  return [...found].filter((name) => name.length > 0 && !SHELL_VARIABLES.has(name));
+}
+function codeOf(path, text, isScript) {
+  const lines = text.split(`
+`);
+  if (isScript)
+    return lines;
+  if (!MARKDOWN.test(path))
+    return [];
+  const out = [];
+  let fenced = false;
+  for (const line of lines) {
+    if (FENCE.test(line)) {
+      fenced = !fenced;
+      continue;
+    }
+    if (fenced)
+      out.push(line);
+    else
+      for (const inline of line.matchAll(/`([^`]+)`/g))
+        out.push(inline[1] ?? "");
+  }
+  return out;
+}
+var FETCH_AND_RUN = [
+  {
+    kind: "pipe to a shell",
+    pattern: /\b(?:curl|wget|iwr|irm|Invoke-WebRequest|Invoke-RestMethod)\b[^|\n]*\|\s*(?:sudo\s+(?:-\S+\s+)*)?(?:sh|bash|zsh|dash|python3?|node|perl|ruby|iex|Invoke-Expression)\b/i
+  },
+  {
+    kind: "runs a downloaded script",
+    pattern: /\b(?:sh|bash|zsh|source)\s+<\(\s*(?:curl|wget)\b|\b(?:sh|bash|zsh)\s+-c\s+["']?\$\(\s*(?:curl|wget)\b|\b(?:iex|Invoke-Expression)\b[^\n]*\b(?:irm|iwr|Invoke-WebRequest|Invoke-RestMethod|DownloadString)\b|\bdeno\s+run\b[^\n]*\bhttps?:\/\//i
+  },
+  {
+    kind: "installs from a URL",
+    pattern: /\b(?:pip3?|pipx|uv\s+pip|npm|pnpm|yarn|bun|gem)\s+(?:install|add|i)\b[^\n|;&]*\s(?:https?:\/\/|git\+|github:)/i
+  },
+  {
+    kind: "runs a package fetched at run time",
+    pattern: /\bnpx\s+(?:-y|--yes)\b|\b(?:pnpm|yarn)\s+dlx\s+\S|\buvx\s+[a-z]|\bpipx\s+run\s+[a-z]/i
+  }
+];
+var URL_HOST = /\bhttps?:\/\/([a-z0-9](?:[a-z0-9-]*[a-z0-9])?(?:\.[a-z0-9](?:[a-z0-9-]*[a-z0-9])?)+)/gi;
+var PLACEHOLDER_HOST = /(^|\.)(example\.(com|org|net)|localhost|local|test|invalid|example|internal)$|^(127\.\d+\.\d+\.\d+|0\.0\.0\.0)$|^(www\.)?(your|my)-?(domain|site|server|host|app|api|company|website)\./;
+var FILE_LIKE_HOST = /\.(jpe?g|png|gif|webp|svg|mp[34]|wav|mov|webm|pdf|md|json|txt|html?|js|ts|py|sh)$/;
+function domainsIn(text) {
+  const out = [];
+  for (const match of text.matchAll(URL_HOST)) {
+    const host = (match[1] ?? "").toLowerCase();
+    if (!/\.[a-z]{2,}$/.test(host) && !/^\d+\.\d+\.\d+\.\d+$/.test(host))
+      continue;
+    if (PLACEHOLDER_HOST.test(host) || FILE_LIKE_HOST.test(host))
+      continue;
+    out.push(host);
+  }
+  return out;
+}
+var INVISIBLE = /[​-‍⁠-⁤᠎‪-‮⁦-⁩]|\uDB40[\uDC00-\uDDEF]/g;
+var PICTOGRAPH = /\p{Extended_Pictographic}/u;
+function codePoint(point) {
+  return `U+${point.toString(16).toUpperCase().padStart(4, "0")}`;
+}
+function invisibleIn(line, lineIndex) {
+  const counts = new Map;
+  const spelled = [];
+  for (const match of line.matchAll(INVISIBLE)) {
+    const char = match[0];
+    const at = match.index;
+    if (lineIndex === 0 && at === 0 && char === "\uFEFF")
+      continue;
+    if (char === "‍") {
+      const before = [...line.slice(0, at)].at(-1) ?? "";
+      const after = [...line.slice(at + 1)][0] ?? "";
+      if (PICTOGRAPH.test(before) && PICTOGRAPH.test(after))
+        continue;
+    }
+    const point = char.codePointAt(0) ?? 0;
+    if (point >= 917536 && point <= 917630)
+      spelled.push(String.fromCharCode(point - 917504));
+    const name = point >= 917504 && point <= 917631 ? "Unicode tag" : codePoint(point);
+    counts.set(name, (counts.get(name) ?? 0) + 1);
+  }
+  if (counts.size === 0)
+    return null;
+  const listed = [...counts.entries()].map(([name, n]) => `${n} × ${name}`).join(", ");
+  return spelled.length === 0 ? listed : `${listed}, spelling "${cap(spelled.join(""))}"`;
+}
+var BASE64_RUN = /[A-Za-z0-9+/]{100,}={0,2}/g;
+var DATA_URI = /data:([\w.+-]+\/[\w.+-]+)?(?:;[\w=-]+)*;base64,([A-Za-z0-9+/]{100,}={0,2})/g;
+var URL_TEXT = /\bhttps?:\/\/\S+/g;
+function encodedIn(line) {
+  const out = [];
+  const withoutUris = line.replace(DATA_URI, (_whole, type, body) => {
+    out.push(`data URI${type === undefined ? "" : ` ${type}`}, ${body.length} characters of base64`);
+    return " ";
+  });
+  for (const match of withoutUris.replace(URL_TEXT, " ").matchAll(BASE64_RUN)) {
+    const run = match[0];
+    if (/[A-Z]/.test(run) && /[a-z]/.test(run) && /[0-9]/.test(run))
+      out.push(`${run.length} characters of base64`);
+  }
+  return out;
+}
+var IGNORE_INSTRUCTIONS = /\b(?:ignore|disregard|forget|override)\b[^.\n]{0,40}?\b(?:previous|prior|above|earlier|preceding|all|any|other|system|your)\b[^.\n]{0,25}?\b(?:instructions?|prompts?|rules|guidelines|directions)\b/i;
+var KEEP_FROM_PERSON = /\b(?:do not|don't|never)\s+(?:tell|inform|notify|alert|(?:mention|reveal)(?:\s+(?:this|it))?\s+to)\s+(?:the\s+)?(?:user|human|person|operator)\b|\bwithout\s+(?:telling|informing|notifying|alerting)\s+(?:the\s+)?(?:user|human|person|operator)\b/i;
+function unquoted(line) {
+  return line.replace(/"[^"]*"|\u201C[^\u201D]*\u201D|\u300C[^\u300D]*\u300D|`[^`]*`/g, " ");
+}
+var COMMENT_WORDS = 4;
+function hiddenInMarkdown(path, text) {
+  const out = [];
+  const push = (kind, line, value) => out.push({ kind, file: path, line, text: value });
+  let fenced = false;
+  let comment = null;
+  for (const [index, line] of text.split(`
+`).entries()) {
+    const invisible = invisibleIn(line, index);
+    if (invisible !== null)
+      push("hidden characters", index + 1, invisible);
+    for (const encoded of encodedIn(line))
+      push("encoded block", index + 1, encoded);
+    const visible = line.replace(INVISIBLE, "");
+    if (IGNORE_INSTRUCTIONS.test(unquoted(visible)))
+      push("says to ignore instructions", index + 1, cap(visible));
+    if (KEEP_FROM_PERSON.test(unquoted(visible)))
+      push("says not to tell the person", index + 1, cap(visible));
+    if (comment === null && FENCE.test(line)) {
+      fenced = !fenced;
+      continue;
+    }
+    if (fenced)
+      continue;
+    let rest = visible;
+    while (rest.length > 0) {
+      if (comment === null) {
+        const open = rest.indexOf("<!--");
+        if (open === -1)
+          break;
+        comment = { start: index + 1, text: [] };
+        rest = rest.slice(open + 4);
+      }
+      const close = rest.indexOf("-->");
+      if (close === -1) {
+        comment.text.push(rest);
+        break;
+      }
+      comment.text.push(rest.slice(0, close));
+      const inner = comment.text.join(" ").replace(/\s+/g, " ").trim();
+      if (inner.split(" ").filter((word) => /[\p{L}\p{N}]/u.test(word)).length >= COMMENT_WORDS) {
+        push("HTML comment", comment.start, cap(inner));
+      }
+      comment = null;
+      rest = rest.slice(close + 3);
+    }
+  }
+  if (comment !== null)
+    push("HTML comment", comment.start, cap(`(never closed) ${comment.text.join(" ")}`));
+  return out;
+}
+function bounded(name, items, cut) {
+  if (items.length <= CONTENTS_LIMITS2.items)
+    return [...items];
+  cut.push(name);
+  return items.slice(0, CONTENTS_LIMITS2.items);
+}
+function escapeRegExp(text) {
+  return text.replace(/[.*+?^${}()|[\]\\]/g, "\\$&");
+}
+function namesFile(markdown, path) {
+  const base = path.slice(path.lastIndexOf("/") + 1);
+  return markdown.includes(path) || new RegExp(`(^|[\\s/\`'"(])${escapeRegExp(base)}($|[\\s\`'")]|[.,:;](\\s|$))`, "m").test(markdown);
+}
+function readSkillContents2(input) {
+  const sorted = [...input].sort((a, b) => compare(a.path, b.path));
+  const skipped = [];
+  const files = [];
+  for (const file of sorted) {
+    if (files.length >= CONTENTS_LIMITS2.files)
+      skipped.push({ path: file.path, why: "over the file limit" });
+    else if (file.text === null)
+      skipped.push({ path: file.path, why: file.notFetched === true ? "not fetched" : "binary" });
+    else if (file.text.length > CONTENTS_LIMITS2.fileChars)
+      skipped.push({ path: file.path, why: "too large" });
+    else
+      files.push({ path: file.path, text: file.text.replace(/\r\n?/g, `
+`) });
+  }
+  const skillMd = files.find((file) => file.path === "SKILL.md")?.text ?? "";
+  const programs = new Map;
+  const add = (name, from) => {
+    if (!programs.has(name))
+      programs.set(name, from);
+  };
+  const requires = readRequires2(skillMd);
+  for (const name of requires.cli)
+    add(name, "declared");
+  for (const name of requires.runs)
+    add(name, "runs");
+  const packages = new Map;
+  const settings = new Set;
+  const domains = new Set;
+  const fetchAndRun = [];
+  const hidden = [];
+  const scripts = skipped.filter((file) => file.why === "not fetched" && SCRIPT.test(file.path)).map((file) => ({ path: file.path, namedInSkillMd: namesFile(skillMd, file.path) }));
+  for (const file of files) {
+    const lines = file.text.split(`
+`);
+    const isScript = SCRIPT.test(file.path) || lines[0]?.startsWith("#!") === true;
+    const code = codeOf(file.path, file.text, isScript);
+    const shell = isShell(file.path, file.text);
+    for (const line of code) {
+      const installed = packagesIn(line);
+      for (const pkg of installed)
+        packages.set(`${pkg.manager} ${pkg.name}`, pkg);
+      if (!isScript)
+        continue;
+      for (const name of programsInScriptLine(line, shell)) {
+        if (!installed.some((pkg) => pkg.name === name))
+          add(name, "script");
+      }
+    }
+    if (isScript) {
+      scripts.push({ path: file.path, namedInSkillMd: namesFile(skillMd, file.path) });
+      const interpreter = interpreterOf(lines[0] ?? "");
+      if (interpreter !== null)
+        add(interpreter, "interpreter");
+    }
+    if (/(^|\/)requirements[\w.-]*\.txt$/i.test(file.path)) {
+      for (const name of requirementsIn(file.text))
+        packages.set(`pip ${name}`, { manager: "pip", name });
+    }
+    for (const setting of settingsIn(file.path, file.text, code))
+      settings.add(setting);
+    for (const host of domainsIn(file.text))
+      domains.add(host);
+    for (const [index, line] of lines.entries()) {
+      const run = FETCH_AND_RUN.find(({ pattern }) => pattern.test(line));
+      if (run !== undefined)
+        fetchAndRun.push({ kind: run.kind, file: file.path, line: index + 1, text: cap(line) });
+    }
+    if (MARKDOWN.test(file.path))
+      hidden.push(...hiddenInMarkdown(file.path, file.text));
+  }
+  const packageNames = new Set([...packages.values()].map((pkg) => pkg.name));
+  for (const [name, from] of programs)
+    if (from !== "declared" && packageNames.has(name))
+      programs.delete(name);
+  const cut = [];
+  return {
+    programs: bounded("programs", [...programs.entries()].map(([name, from]) => ({ name, from })).sort((a, b) => compare(a.name, b.name)), cut),
+    packages: bounded("packages", [...packages.values()].sort((a, b) => compare(`${a.manager} ${a.name}`, `${b.manager} ${b.name}`)), cut),
+    settings: bounded("settings", [...settings].sort(compare), cut),
+    scripts: bounded("scripts", scripts.sort((a, b) => compare(a.path, b.path)), cut),
+    fetchAndRun: bounded("fetchAndRun", fetchAndRun, cut),
+    domains: bounded("domains", [...domains].sort(compare), cut),
+    hidden: bounded("hidden", hidden, cut),
+    read: {
+      files: files.length,
+      bytes: files.reduce((sum, file) => sum + new TextEncoder().encode(file.text).byteLength, 0),
+      skipped
+    },
+    cut
+  };
+}
+function contentsLines2(contents, perList = 6) {
+  const list = (items, cut) => items.length === 0 ? "none found" : `${items.slice(0, perList).join(", ")}${items.length > perList || cut ? `, and ${Math.max(0, items.length - perList)}${cut ? "+" : ""} more` : ""}`;
+  const at = (rows, name) => list(rows.map((row) => `${row.kind} at ${row.file}:${row.line} "${row.text}"`), contents.cut.includes(name));
+  return [
+    `Programs: ${list(contents.programs.map((p) => `${p.name} (${p.from})`), contents.cut.includes("programs"))}`,
+    ...contents.packages.length === 0 ? [] : [`Packages: ${list(contents.packages.map((p) => `${p.manager} ${p.name}`), contents.cut.includes("packages"))}`],
+    `Keys and settings it reads: ${list(contents.settings, contents.cut.includes("settings"))}`,
+    `Scripts: ${list(contents.scripts.map((s) => s.namedInSkillMd ? `${s.path} (named in SKILL.md)` : s.path), contents.cut.includes("scripts"))}`,
+    `Fetches and runs code: ${at(contents.fetchAndRun, "fetchAndRun")}`,
+    `Domains named: ${list(contents.domains, contents.cut.includes("domains"))}`,
+    `Hidden or encoded text: ${at(contents.hidden, "hidden")}`,
+    ...contents.read.skipped.length === 0 ? [] : [`Not read: ${list(contents.read.skipped.map((s) => `${s.path} (${s.why})`), false)}`]
+  ];
+}
+
+// ../../packages/protocol/src/pulse.ts
+var MAX_SKILL_FILES_READ2 = 200;
+var MAX_SCRIPTS_READ = 5;
+var GITHUB_READ_TIMEOUT_MS = 1e4;
+var INSTALLER_ORDER = [
+  "",
+  "skills/",
+  "skills/.curated/",
+  "skills/.experimental/",
+  "skills/.system/",
+  ".agents/skills/",
+  ".claude/skills/",
+  ".cline/skills/",
+  ".codebuddy/skills/",
+  ".codex/skills/",
+  ".commandcode/skills/",
+  ".continue/skills/",
+  ".factory/skills/",
+  ".github/skills/",
+  ".goose/skills/",
+  ".grok/skills/",
+  ".iflow/skills/",
+  ".junie/skills/",
+  ".kilo/skills/",
+  ".kilocode/skills/",
+  ".kimchi/skills/",
+  ".kiro/skills/",
+  ".minimax/skills/",
+  ".mux/skills/",
+  ".neovate/skills/",
+  ".opencode/skills/",
+  ".openhands/skills/",
+  ".pi/skills/",
+  ".posit/assistant/skills/",
+  ".qoder/skills/",
+  ".roo/skills/",
+  ".trae/skills/",
+  ".windsurf/skills/",
+  ".zcode/skills/",
+  ".zencoder/skills/"
+];
+function installerRank(path) {
+  let best = null;
+  INSTALLER_ORDER.forEach((prefix, rank) => {
+    if (best !== null || !path.startsWith(prefix))
+      return;
+    const depth = path.slice(prefix.length).split("/").length;
+    if (prefix === "" ? depth <= 2 : depth <= 4)
+      best = rank;
+  });
+  return best;
+}
+function normaliseName(name) {
+  return name.trim().toLowerCase().replace(/[\s_]+/g, "-");
+}
+function frontmatterName2(markdown) {
+  const lines = markdown.split(`
+`);
+  if (lines[0]?.trim() !== "---")
+    return null;
+  for (const line of lines.slice(1)) {
+    if (line.trim() === "---")
+      return null;
+    const match = /^name\s*:\s*(.+?)\s*$/.exec(line);
+    if (match !== null)
+      return (match[1] ?? "").replace(/^["']|["']$/g, "");
+  }
+  return null;
+}
+var KEY_NAME3 = /\b[A-Z][A-Z0-9_]*(?:API_KEY|APIKEY|_TOKEN|_SECRET)\b/g;
+var PAID_WORD = /\b(pricing|paid|subscription|credits?|billing|free tier|per minute)\b/gi;
+var HOST = /https?:\/\/([a-zA-Z0-9.-]+\.[a-zA-Z]{2,})/g;
+var NOT_A_SERVICE = /(^|\.)(github\.com|githubusercontent\.com|example\.(com|org|net))$/;
+var LOOKS_LIKE_A_FILE = /\.(jpe?g|png|gif|webp|mp[34]|wav|mov|webm|pdf|md|json|txt)$/;
+var SCRIPT2 = /\.(sh|bash|py|js|mjs|cjs|ts|rb|pl)$/i;
+var RUNS_FILE = /\b(?:node|bun|deno|tsx|python3?|bash|sh)\s+((?:~|\.{1,2})?\/?[\w.@-]+(?:\/[\w.@-]+)*\.(?:mjs|cjs|js|ts|py|sh))\b/g;
+var PLACEHOLDER = /(^|\/)(path|your|my|foo|example|script|file|input|output)(\/|\.|$)/i;
+function filesItRuns(markdown, repositoryFiles) {
+  const named = [...markdown.matchAll(RUNS_FILE)].map((match) => match[1] ?? "");
+  const found = new Set;
+  for (const name of named) {
+    if (name.startsWith("/") || name.startsWith("~") || PLACEHOLDER.test(name))
+      continue;
+    const tail = name.replace(/^(\.{1,2}\/)+/, "");
+    const hit = repositoryFiles.find((file) => file.path === tail || file.path.endsWith(`/${tail}`));
+    if (hit !== undefined)
+      found.add(hit.path);
+  }
+  return [...found];
+}
+function absentFiles(markdown, repositoryFiles) {
+  const named = [...markdown.matchAll(RUNS_FILE)].map((match) => match[1] ?? "");
+  return [...new Set(named)].filter((name) => {
+    if (name.startsWith("/") || name.startsWith("~") || PLACEHOLDER.test(name))
+      return false;
+    const tail = name.replace(/^(\.{1,2}\/)+/, "");
+    return !repositoryFiles.some((file) => file.path === tail || file.path.endsWith(`/${tail}`));
+  });
+}
+function readSkillFacts2(markdown, filesInDirectory, repositoryFiles = filesInDirectory) {
+  const hosts = [...markdown.matchAll(HOST)].map((match) => (match[1] ?? "").toLowerCase()).filter((host) => !NOT_A_SERVICE.test(host) && !LOOKS_LIKE_A_FILE.test(host));
+  return {
+    files: filesInDirectory.length,
+    bytes: filesInDirectory.reduce((total, file) => total + file.size, 0),
+    scripts: filesInDirectory.filter((file) => SCRIPT2.test(file.path)).length,
+    keys: [...new Set(markdown.match(KEY_NAME3) ?? [])],
+    paidWords: [...new Set((markdown.match(PAID_WORD) ?? []).map((word) => word.toLowerCase()))],
+    hosts: [...new Set(hosts)],
+    runsAbsentFiles: absentFiles(markdown, repositoryFiles),
+    keysInScriptsItRuns: [],
+    requires: readRequires2(markdown),
+    contains: readSkillContents2([{ path: "SKILL.md", text: markdown }])
+  };
+}
+function parentOf(path) {
+  const at = path.lastIndexOf("/");
+  return at === -1 ? "" : path.slice(0, at);
+}
+async function locate(source, repo, skill, listing) {
+  const wanted = normaliseName(skill);
+  const candidates = listing.files.filter((file) => skillFilesIn2([file]).length === 1);
+  const guessed = candidates.filter((file) => normaliseName(directoryOf2(file.path)) === wanted);
+  const others = candidates.filter((file) => !guessed.includes(file));
+  const read = async (files) => (await Promise.all(files.map(async (file) => ({ file, markdown: await source.file(repo, file.path) })))).filter((entry) => entry.markdown !== null && normaliseName(frontmatterName2(entry.markdown) ?? "") === wanted);
+  let named = await read(guessed);
+  if (named.length === 0) {
+    if (others.length > MAX_SKILL_FILES_READ2) {
+      return { status: "incomplete", reason: "too_many_skill_files" };
+    }
+    for (let at = 0;at < others.length && named.length === 0; at += 20) {
+      named = await read(others.slice(at, at + 20));
+    }
+    if (named.length > 0) {
+      const last = others.findIndex((file) => file === named.at(-1)?.file);
+      named = [...named, ...await read(others.slice(last + 1))];
+    }
+  }
+  if (named.length === 0) {
+    if (listing.truncated)
+      return { status: "incomplete", reason: "truncated_listing" };
+    return { status: "absent", skillFiles: candidates.length };
+  }
+  const ranked = named.map((entry) => ({ ...entry, rank: installerRank(entry.file.path) })).sort((a, b) => (a.rank ?? Infinity) - (b.rank ?? Infinity));
+  const [first2, ...rest] = ranked;
+  if (first2 === undefined)
+    return { status: "absent", skillFiles: candidates.length };
+  const differs = rest.filter((entry) => entry.file.sha !== first2.file.sha);
+  if (first2.rank === null && differs.length > 0) {
+    return { status: "ambiguous", paths: ranked.map((entry) => entry.file.path) };
+  }
+  return {
+    status: "found",
+    file: first2.file,
+    markdown: first2.markdown,
+    alsoAt: rest.filter((entry) => entry.file.sha === first2.file.sha).map((entry) => entry.file.path),
+    differsAt: differs.map((entry) => entry.file.path)
+  };
+}
+function usualPaths(skill) {
+  return [`skills/${skill}/SKILL.md`, `${skill}/SKILL.md`, `.claude/skills/${skill}/SKILL.md`, "SKILL.md"];
+}
+async function readFileOnly(result, source) {
+  if (!/^[\w.-]+$/.test(result.skill) || /^\.+$/.test(result.skill))
+    return null;
+  const wanted = normaliseName(result.skill);
+  for (const path of usualPaths(result.skill)) {
+    const markdown = await source.file(result.source, path);
+    if (markdown === null || normaliseName(frontmatterName2(markdown) ?? "") !== wanted)
+      continue;
+    const { keys, paidWords, hosts, requires, contains } = readSkillFacts2(markdown, []);
+    return { ...result, status: "read_file_only", path, keys, paidWords, hosts, requires, contains };
+  }
+  return null;
+}
+async function readPulse2(result, source) {
+  if (!/^(?!\.{1,2}\/)[\w.-]+\/(?!\.{1,2}$)[\w.-]+$/.test(result.source))
+    return { ...result, status: "not_github" };
+  const listing = await source.listing(result.source);
+  if (listing === "rate_limited")
+    return await readFileOnly(result, source) ?? { ...result, status: "rate_limited" };
+  if (listing === "unreachable")
+    return { ...result, status: "repository_unreachable" };
+  const located = await locate(source, result.source, result.skill, listing);
+  if (located.status === "absent")
+    return { ...result, status: "absent", skillFiles: located.skillFiles };
+  if (located.status === "ambiguous")
+    return { ...result, status: "ambiguous", paths: located.paths };
+  if (located.status === "incomplete")
+    return { ...result, status: "incomplete", reason: located.reason };
+  const directory = parentOf(located.file.path);
+  const prefix = directory === "" ? "" : `${directory}/`;
+  const inDirectory = listing.files.filter((file) => file.path.startsWith(prefix));
+  const scripts = filesItRuns(located.markdown, listing.files).slice(0, MAX_SCRIPTS_READ);
+  const scriptTexts = await Promise.all(scripts.map(async (file) => ({ file, text: await source.file(result.source, file) })));
+  const keysInScriptsItRuns = scriptTexts.map(({ file, text }) => ({ file, keys: [...new Set(text?.match(KEY_NAME3) ?? [])] })).filter((entry) => entry.keys.length > 0);
+  const relative = (path) => path.startsWith(prefix) ? path.slice(prefix.length) : path;
+  const fetched = new Set(scriptTexts.filter((entry) => entry.text !== null).map((entry) => entry.file));
+  const contains = readSkillContents2([
+    { path: "SKILL.md", text: located.markdown },
+    ...scriptTexts.flatMap(({ file, text }) => text === null ? [] : [{ path: relative(file), text }]),
+    ...inDirectory.filter((file) => SCRIPT2.test(file.path) && !fetched.has(file.path)).map((file) => ({ path: relative(file.path), text: null, notFetched: true }))
+  ]);
+  return {
+    ...result,
+    status: "read",
+    path: located.file.path,
+    alsoAt: located.alsoAt,
+    differsAt: located.differsAt,
+    contentId: directory === "" ? listing.rootTreeId : listing.trees.get(directory) ?? located.file.sha,
+    skillMdSha: located.file.sha,
+    repository: await source.repository(result.source),
+    ...listing.readAt === undefined ? {} : { listingReadAt: listing.readAt },
+    ...readSkillFacts2(located.markdown, inDirectory, listing.files),
+    keysInScriptsItRuns,
+    contains
+  };
+}
+function githubPulseSource2(options = {}) {
+  const get = options.fetch ?? fetch;
+  const timeoutMs = options.timeoutMs ?? GITHUB_READ_TIMEOUT_MS;
+  const bounded2 = () => {
+    const limit = AbortSignal.timeout(timeoutMs);
+    return options.signal === undefined ? limit : AbortSignal.any([limit, options.signal]);
+  };
+  const headers = {
+    accept: "application/vnd.github+json",
+    "user-agent": "arp-pulse",
+    ...options.token === undefined ? {} : { authorization: `Bearer ${options.token}` }
+  };
+  const listings = new Map;
+  const repositories = new Map;
+  const quota = { apiRequests: 0, limited: false, resetsAt: null };
+  const noteLimit = (response) => {
+    const hit = response.status === 429 || response.status === 403 && (response.headers.get("x-ratelimit-remaining") === "0" || response.headers.has("retry-after"));
+    if (!hit)
+      return false;
+    quota.limited = true;
+    const reset = Number(response.headers.get("x-ratelimit-reset"));
+    if (Number.isFinite(reset) && reset > 0)
+      quota.resetsAt = new Date(reset * 1000).toISOString();
+    return true;
+  };
+  const readListing = async (repo) => {
+    const cached = await options.cache?.get(repo).catch(() => null);
+    if (cached !== null && cached !== undefined)
+      return cached;
+    if (quota.limited)
+      return "rate_limited";
+    quota.apiRequests += 1;
+    const response = await get(`https://api.github.com/repos/${repo}/git/trees/HEAD?recursive=1`, { headers, signal: bounded2() });
+    if (noteLimit(response))
+      return "rate_limited";
+    if (!response.ok)
+      return "unreachable";
+    const body = await response.json();
+    const files = [];
+    const trees = new Map;
+    for (const entry of body.tree ?? []) {
+      if (typeof entry.path !== "string" || typeof entry.sha !== "string")
+        continue;
+      if (entry.type === "tree")
+        trees.set(entry.path, entry.sha);
+      if (entry.type === "blob") {
+        files.push({ path: entry.path, sha: entry.sha, size: typeof entry.size === "number" ? entry.size : 0 });
+      }
+    }
+    const listing = {
+      rootTreeId: typeof body.sha === "string" ? body.sha : "",
+      files,
+      trees,
+      truncated: body.truncated === true
+    };
+    await options.cache?.set(repo, listing).catch(() => {
+      return;
+    });
+    return listing;
+  };
+  const readRepository = async (repo) => {
+    if (options.withRepository === false || quota.limited)
+      return null;
+    quota.apiRequests += 1;
+    const response = await get(`https://api.github.com/repos/${repo}`, { headers, signal: bounded2() });
+    noteLimit(response);
+    if (!response.ok)
+      return null;
+    const body = await response.json();
+    return {
+      stars: typeof body.stargazers_count === "number" ? body.stargazers_count : 0,
+      pushedAt: typeof body.pushed_at === "string" ? body.pushed_at : null,
+      archived: body.archived === true
+    };
+  };
+  const once = (cache, repo, read) => {
+    const hit = cache.get(repo);
+    if (hit !== undefined)
+      return hit;
+    const started = read(repo);
+    cache.set(repo, started);
+    return started;
+  };
+  return {
+    listing: (repo) => once(listings, repo, readListing),
+    repository: (repo) => once(repositories, repo, readRepository),
+    file: async (repo, path) => {
+      const encoded = path.split("/").map(encodeURIComponent).join("/");
+      const response = await get(`https://raw.githubusercontent.com/${repo}/HEAD/${encoded}`, { signal: bounded2() });
+      return response.ok ? response.text() : null;
+    },
+    quota: () => ({ ...quota })
+  };
+}
+
+// ../../packages/protocol/src/skill-install.ts
+var MAX_SKILLS_READ2 = 3;
+var ADD = new Set(["add", "a", "install", "i"]);
+var RUNNERS = new Set(["npx", "pnpx", "bunx"]);
+var DLX = { pnpm: ["dlx"], yarn: ["dlx"], npm: ["exec", "x"] };
+var SKILLS = /^skills(@[\w.^~<>=*-]+)?$/;
+var ENV_ASSIGNMENT = /^[A-Za-z_][A-Za-z0-9_]*=/;
+function shellWords(command) {
+  const commands = [[]];
+  let word = null;
+  let quote = null;
+  const end = () => {
+    if (word !== null)
+      commands[commands.length - 1]?.push(word);
+    word = null;
+  };
+  for (let at = 0;at < command.length; at += 1) {
+    const char = command[at] ?? "";
+    if (quote !== null) {
+      if (char === quote)
+        quote = null;
+      else if (char === "\\" && quote === '"' && at + 1 < command.length)
+        word = (word ?? "") + command[at += 1];
+      else
+        word = (word ?? "") + char;
+    } else if (char === "'" || char === '"') {
+      quote = char;
+      word ??= "";
+    } else if (char === "\\" && at + 1 < command.length) {
+      word = (word ?? "") + command[at += 1];
+    } else if (/\s/.test(char) && char !== `
+`) {
+      end();
+    } else if (char === `
+` || char === ";" || char === "&" || char === "|") {
+      end();
+      commands.push([]);
+    } else {
+      word = (word ?? "") + char;
+    }
+  }
+  end();
+  return commands.filter((words) => words.length > 0);
+}
+function addArguments(words) {
+  let at = 0;
+  while (at < words.length && ENV_ASSIGNMENT.test(words[at] ?? ""))
+    at += 1;
+  const first2 = words[at];
+  if (first2 !== undefined && RUNNERS.has(first2))
+    at += 1;
+  else if (first2 !== undefined && DLX[first2]?.includes(words[at + 1] ?? "") === true)
+    at += 2;
+  while (at < words.length && (words[at] ?? "").startsWith("-"))
+    at += 1;
+  if (!SKILLS.test(words[at] ?? ""))
+    return null;
+  const subcommand = words[at + 1];
+  return subcommand !== undefined && ADD.has(subcommand) ? words.slice(at + 2) : null;
+}
+function parseAdd(args) {
+  const skills = [];
+  let source = null;
+  let all = false;
+  let list = false;
+  for (let at = 0;at < args.length; at += 1) {
+    const arg = args[at] ?? "";
+    if (arg === "--all")
+      all = true;
+    else if (arg === "-l" || arg === "--list")
+      list = true;
+    else if (arg === "-s" || arg === "--skill" || arg === "-a" || arg === "--agent" || arg === "--subagent") {
+      while (at + 1 < args.length && !(args[at + 1] ?? "").startsWith("-")) {
+        at += 1;
+        if (arg === "-s" || arg === "--skill")
+          skills.push(args[at] ?? "");
+      }
+    } else if (arg === "--metadata")
+      at += 1;
+    else if (!arg.startsWith("-") && source === null)
+      source = arg;
+  }
+  return { source, skills, all, list };
+}
+var NAME = /^[\w.-]+$/;
+function github(repo, ref, subpath, named, all) {
+  const [owner, name] = repo.split("/");
+  const clean = `${owner}/${(name ?? "").replace(/\.git$/, "")}`;
+  const path = subpath?.replace(/^\/+|\/+$/g, "") ?? "";
+  return {
+    kind: "github",
+    repo: clean,
+    ref,
+    subpath: path === "" || path.split("/").includes("..") ? null : path,
+    skills: all || named.length === 0 || named.includes("*") ? "all" : [...new Set(named)]
+  };
+}
+function sourceOf(input, named, all) {
+  const elsewhere = { kind: "elsewhere", source: input };
+  if (/^(\.{1,2}(\/|$)|\/|~)/.test(input) || /^[A-Za-z]:[\\/]/.test(input))
+    return elsewhere;
+  const hash = input.indexOf("#");
+  let rest = hash === -1 ? input : input.slice(0, hash);
+  const fragment = hash === -1 ? "" : input.slice(hash + 1);
+  const fragmentAt = fragment.indexOf("@");
+  let ref = (fragmentAt === -1 ? fragment : fragment.slice(0, fragmentAt)) || null;
+  const fragmentSkill = fragmentAt === -1 ? null : fragment.slice(fragmentAt + 1) || null;
+  const withSkill = (names) => fragmentSkill === null ? names : [fragmentSkill];
+  const prefixed = /^github:(.+)$/.exec(rest);
+  if (prefixed?.[1] !== undefined)
+    rest = prefixed[1];
+  const tree = /^https?:\/\/(?:www\.)?github\.com\/([^/]+)\/([^/]+)\/tree\/([^/]+)(?:\/(.+))?$/.exec(rest);
+  if (tree !== null) {
+    ref = tree[3] ?? ref;
+    return github(`${tree[1]}/${tree[2]}`, ref, tree[4] ?? null, withSkill(named), all);
+  }
+  const url = /^https?:\/\/(?:www\.)?github\.com\/([^/]+)\/([^/]+?)\/?$/.exec(rest);
+  if (url !== null)
+    return github(`${url[1]}/${url[2]}`, ref, null, withSkill(named), all);
+  if (rest.includes(":"))
+    return elsewhere;
+  const at = /^([^/]+)\/([^/@]+)@(.+)$/.exec(rest);
+  if (at !== null)
+    return github(`${at[1]}/${at[2]}`, ref, null, [fragmentSkill ?? at[3] ?? ""], false);
+  const shorthand = /^([^/]+)\/([^/]+?)(?:\/(.+?))?\/?$/.exec(rest);
+  if (shorthand !== null && NAME.test(shorthand[1] ?? "") && NAME.test(shorthand[2] ?? "")) {
+    return github(`${shorthand[1]}/${shorthand[2]}`, ref, shorthand[3] ?? null, withSkill(named), all);
+  }
+  return elsewhere;
+}
+function skillInstallOf2(command) {
+  for (const words of shellWords(command)) {
+    const args = addArguments(words);
+    if (args === null)
+      continue;
+    const add = parseAdd(args);
+    if (add.list || add.source === null)
+      continue;
+    return sourceOf(add.source, add.skills, add.all);
+  }
+  return null;
+}
+async function namesUnder(install, source) {
+  const listing = await source.listing(install.repo);
+  if (listing === "rate_limited" || listing === "unreachable")
+    return listing;
+  const prefix = install.subpath === null ? "" : `${install.subpath}/`;
+  const files = skillFilesIn2(listing.files).filter((file) => file.path.startsWith(prefix)).sort((a, b) => a.path.split("/").length - b.path.split("/").length || a.path.localeCompare(b.path));
+  const names = await Promise.all(files.slice(0, MAX_SKILLS_READ2).map(async (file) => {
+    const markdown = await source.file(install.repo, file.path);
+    const parts = file.path.split("/");
+    return (markdown === null ? null : frontmatterName2(markdown)) ?? parts.at(-2) ?? install.repo.split("/")[1] ?? "";
+  }));
+  return { names, total: files.length };
+}
+async function readSkillInstall2(install, source, deadline) {
+  if (install.kind === "elsewhere")
+    return install;
+  const late = deadline.then(() => "late");
+  let names;
+  let total;
+  if (install.skills === "all") {
+    const found = await Promise.race([
+      namesUnder(install, source).catch(() => "unreachable"),
+      late
+    ]);
+    if (found === "late")
+      return {
+        kind: "github",
+        install,
+        readings: [],
+        notRead: 0,
+        listing: "no answer in time"
+      };
+    if (found === "rate_limited" || found === "unreachable") {
+      return { kind: "github", install, readings: [], notRead: 0, listing: found };
+    }
+    ({ names, total } = found);
+  } else {
+    names = install.skills.slice(0, MAX_SKILLS_READ2);
+    total = install.skills.length;
+  }
+  const readings = await Promise.all(names.map(async (skill) => {
+    const pulse = readPulse2({ source: install.repo, skill, installs: null, installsText: null, url: null }, source).then((read) => ({ skill, pulse: read })).catch(() => ({ skill, pulse: null, why: "read failed" }));
+    const settled = await Promise.race([pulse, late]);
+    return settled === "late" ? { skill, pulse: null, why: "no answer in time" } : settled;
+  }));
+  return {
+    kind: "github",
+    install,
+    readings,
+    notRead: Math.max(0, total - names.length),
+    listing: null
+  };
+}
+var plural = (count, word) => `${count} ${word}${count === 1 ? "" : "s"}`;
+function readingLines(repo, reading) {
+  const head = `${repo} · ${reading.skill}`;
+  const { pulse } = reading;
+  if (pulse === null)
+    return [
+      `${head}: not read (${reading.why === "no answer in time" ? "GitHub did not answer in time" : "the read failed"})`
+    ];
+  switch (pulse.status) {
+    case "read": {
+      const facts = [
+        `${pulse.path}`,
+        plural(pulse.files, "file"),
+        ...pulse.repository === null ? [] : [
+          plural(pulse.repository.stars, "star"),
+          ...pulse.repository.pushedAt === null ? [] : [`last push ${pulse.repository.pushedAt.slice(0, 10)}`],
+          ...pulse.repository.archived ? ["archived"] : []
+        ]
+      ];
+      return [
+        `${head} (${facts.join(", ")})`,
+        ...contentsLines2(pulse.contains).map((line) => `  ${line}`),
+        ...pulse.differsAt.length === 0 ? [] : [`  Other files carry this name and differ: ${pulse.differsAt.join(", ")}`]
+      ];
+    }
+    case "read_file_only":
+      return [
+        `${head} (${pulse.path}; GitHub's hourly limit was reached, so SKILL.md alone was read, not its scripts)`,
+        ...contentsLines2(pulse.contains).map((line) => `  ${line}`)
+      ];
+    case "absent":
+      return [
+        `${head}: no SKILL.md in the repository names it (${plural(pulse.skillFiles, "skill file")} read)`
+      ];
+    case "ambiguous":
+      return [
+        `${head}: several files carry this name and differ, and which one installs is not knowable here: ${pulse.paths.join(", ")}`
+      ];
+    case "incomplete":
+      return [
+        `${head}: not read (${pulse.reason === "truncated_listing" ? "GitHub cut the repository's file list" : "too many skill files to search"})`
+      ];
+    case "rate_limited":
+      return [`${head}: not read (GitHub's hourly limit; GITHUB_TOKEN raises it)`];
+    case "repository_unreachable":
+      return [`${head}: not read (the repository did not answer)`];
+    case "not_github":
+      return [`${head}: not read (not a GitHub repository)`];
+  }
+}
+function skillInstallReport2(read) {
+  const title = "ARP read this skill before the install (what its own files contain; facts, not a verdict):";
+  if (read.kind === "elsewhere") {
+    return `${title}
+${read.source}: not read (ARP reads GitHub sources only)`;
+  }
+  const { install } = read;
+  const lines = [title];
+  if (read.listing !== null) {
+    const why = read.listing === "rate_limited" ? "GitHub's hourly limit; GITHUB_TOKEN raises it" : read.listing === "no answer in time" ? "GitHub did not answer in time" : "the repository did not answer";
+    lines.push(`${install.repo}${install.subpath === null ? "" : `/${install.subpath}`}: not read (${why})`);
+  } else if (read.readings.length === 0) {
+    lines.push(`${install.repo}${install.subpath === null ? "" : `/${install.subpath}`}: no SKILL.md found`);
+  }
+  for (const reading of read.readings)
+    lines.push(...readingLines(install.repo, reading));
+  if (read.notRead > 0)
+    lines.push(`${plural(read.notRead, "more skill")} in this install not read (ARP reads ${MAX_SKILLS_READ2})`);
+  if (install.ref !== null)
+    lines.push(`Read at the default branch's HEAD; the command names ${install.ref}.`);
+  return lines.join(`
+`);
+}
+
+// ../hook/src/pre-install.ts
+var GITHUB_BUDGET_MS = 6000;
+var OUTPUT_CAP = 9500;
+var FOR_THE_AGENT = "This is what the skill's own files contain, read by ARP's hook before the install ran. " + "They are facts, not a verdict, and nothing was blocked. The person was shown the same lines.";
+function capped(text) {
+  return text.length <= OUTPUT_CAP ? text : `${text.slice(0, OUTPUT_CAP)}
+… (cut)`;
+}
+async function preToolUse(raw, options = {}) {
+  const call = toolCallOf(raw);
+  if (call === null || call.tool !== "Bash")
+    return null;
+  const command = call.input?.["command"];
+  if (typeof command !== "string")
+    return null;
+  const install = skillInstallOf2(command);
+  if (install === null)
+    return null;
+  const env = options.env ?? process.env;
+  const token = (env["GITHUB_TOKEN"] ?? env["GH_TOKEN"])?.trim() || undefined;
+  const budgetMs = options.budgetMs ?? GITHUB_BUDGET_MS;
+  const deadline = new Promise((resolve) => setTimeout(resolve, budgetMs));
+  const source = githubPulseSource2({
+    ...options.fetch === undefined ? {} : { fetch: options.fetch },
+    ...token === undefined ? {} : { token },
+    timeoutMs: budgetMs + 500,
+    signal: AbortSignal.timeout(budgetMs + 500)
+  });
+  const report = capped(skillInstallReport2(await readSkillInstall2(install, source, deadline)));
+  return JSON.stringify({
+    systemMessage: report,
+    hookSpecificOutput: {
+      hookEventName: "PreToolUse",
+      additionalContext: `${report}
+
+${FOR_THE_AGENT}`
+    }
+  });
+}
+
+// ../hook/src/failure-answer.ts
+import { createHash } from "node:crypto";
+
+// ../../packages/protocol/src/declare/line-need.ts
+var NEEDS_LABEL3 = "needs";
+var WITH_LABEL3 = "with";
+var PROGRAM_NAME3 = /^[a-z0-9][a-z0-9._+-]{0,39}$/;
+var KEY_NAME4 = /^[A-Z][A-Z0-9_]{2,63}$/;
+function needKindOf3(name) {
+  if (PROGRAM_NAME3.test(name))
+    return "program";
+  if (KEY_NAME4.test(name))
+    return "key";
+  return null;
+}
+function needKindOfReason3(reason) {
+  if (reason === "missing-program")
+    return "program";
+  if (reason === "missing-key")
+    return "key";
+  return null;
+}
+var escaped = (text) => text.replace(/[.*+?^${}()|[\]\\]/g, "\\$&");
+function namedInSkill3(name, markdown) {
+  const kind = needKindOf3(name);
+  if (kind === null)
+    return false;
+  const pattern = new RegExp(`(?<![A-Za-z0-9_+.-])${escaped(name)}(?![A-Za-z0-9_+-])(?!\\.[A-Za-z0-9])`, kind === "program" ? "i" : "");
+  return pattern.test(markdown);
+}
+function publishableNeed3(name, markdown, kind) {
+  const shape = needKindOf3(name);
+  if (shape === null || kind !== undefined && shape !== kind)
+    return null;
+  return namedInSkill3(name, markdown) ? name : null;
+}
+
+// ../../packages/protocol/src/declare/thing-cid.ts
+function thingDocument2(thing) {
+  return JSON.stringify({
+    "@context": "https://schema.org",
+    "@type": "Thing",
+    description: thing.description,
+    image: thing.image,
+    name: thing.name,
+    url: thing.url
+  });
+}
+function base32(bytes) {
+  const alphabet = "abcdefghijklmnopqrstuvwxyz234567";
+  let bits = 0;
+  let value = 0;
+  let out = "";
+  for (const byte of bytes) {
+    value = value << 8 | byte;
+    bits += 8;
+    while (bits >= 5) {
+      out += alphabet[value >>> bits - 5 & 31];
+      bits -= 5;
+    }
+  }
+  if (bits > 0)
+    out += alphabet[value << 5 - bits & 31];
+  return out;
+}
+function thingUriOfDigest(digest) {
+  const cid = new Uint8Array(4 + digest.length);
+  cid.set([1, 85, 18, 32]);
+  cid.set(digest, 4);
+  return `ipfs://b${base32(cid)}`;
+}
+
+// ../../packages/protocol/src/declare/tool-thing.ts
+class InvalidToolUrlError2 extends Error {
+  input;
+  constructor(input, why) {
+    super(`${JSON.stringify(input)} is not a usable tool URL: ${why}`);
+    this.name = "InvalidToolUrlError";
+    this.input = input;
+  }
+}
+var TOOL_THING_DESCRIPTION2 = "A tool identified by its URL, so every declaration of its use lands on one atom.";
+function normaliseToolUrl2(input) {
+  const trimmed = input.trim();
+  if (trimmed.length === 0)
+    throw new InvalidToolUrlError2(input, "empty");
+  const withScheme = /^[a-z][a-z0-9+.-]*:/i.test(trimmed) ? trimmed : `https://${trimmed}`;
+  let url;
+  try {
+    url = new URL(withScheme);
+  } catch {
+    throw new InvalidToolUrlError2(input, "not parseable as a URL");
+  }
+  if (url.protocol !== "http:" && url.protocol !== "https:") {
+    throw new InvalidToolUrlError2(input, `scheme ${url.protocol} is not http or https`);
+  }
+  if (url.hostname.length === 0)
+    throw new InvalidToolUrlError2(input, "no host");
+  const host = url.hostname.replace(/\.+$/, "").replace(/^(?:www\.)+/, "");
+  const path = url.pathname.replace(/(?:\.git|\/)+$/, "");
+  const params = [...new URLSearchParams(url.search).entries()].sort(([a], [b]) => a < b ? -1 : a > b ? 1 : 0);
+  const query = params.length === 0 ? "" : `?${new URLSearchParams(params).toString()}`;
+  const port = url.port === "" || url.port === "80" || url.port === "443" ? "" : `:${url.port}`;
+  return `https://${host}${port}${path}${query}`;
+}
+function toolThing2(url) {
+  const normalised = normaliseToolUrl2(url);
+  return {
+    name: normalised.slice("https://".length),
+    description: TOOL_THING_DESCRIPTION2,
+    image: "",
+    url: normalised
+  };
+}
+
+// ../../packages/protocol/src/declare/usage-words.ts
+var USAGE_OUTCOMES3 = ["worked", "failed"];
+var HAS_OUTCOME_LABEL3 = "has outcome";
+var USAGE_REASONS3 = ["missing-key", "missing-program", "missing-file", "not-the-right-tool", "no-delivery"];
+var BECAUSE_LABEL3 = "because";
+
+// ../../packages/protocol/src/sources/intuition/terms.ts
+var USES2 = "0x5c0bde1cc696456c0268248c4656acdf9621fdb39e605bc99b0a83dc8ff6e800";
+
+// ../../packages/protocol/src/pulse-usage.ts
+var DEFAULT_LIMIT = 250;
+var rowsOf = (data) => Array.isArray(data["triples"]) ? data["triples"].filter(isRecord) : [];
+async function findToolAtoms2(uris, config) {
+  const out = new Map;
+  if (uris.length === 0)
+    return out;
+  const transport = config.transport ?? createGraphqlTransport({
+    url: config.graphqlUrl,
+    fetch: config.fetch ?? defaultFetch(),
+    timeoutMs: config.timeoutMs ?? DEFAULT_TIMEOUT_MS
+  });
+  const data = await transport.request(PULSE_ATOMS_QUERY, { uris: [...uris], limit: uris.length });
+  const rows = Array.isArray(data["atoms"]) ? data["atoms"].filter(isRecord) : [];
+  for (const row of rows) {
+    const id = readString(row, "term_id");
+    const uri = readString(row, "data");
+    if (id !== null && uri !== null)
+      out.set(uri, id);
+  }
+  return out;
+}
+async function getPulseUsage2(toolAtomIds, config) {
+  const transport = config.transport ?? createGraphqlTransport({
+    url: config.graphqlUrl,
+    fetch: config.fetch ?? defaultFetch(),
+    timeoutMs: config.timeoutMs ?? DEFAULT_TIMEOUT_MS
+  });
+  const limit = Math.min(config.limit ?? DEFAULT_LIMIT, DEFAULT_LIMIT);
+  const key = (id) => id.toLowerCase();
+  const empty = () => ({
+    declarations: [],
+    agents: 0,
+    wallets: 0,
+    byDomain: {},
+    byOutcome: {},
+    byReason: {},
+    byNeed: {},
+    byWith: {},
+    byOthers: [],
+    truncated: false
+  });
+  const out = new Map(toolAtomIds.map((id) => [key(id), empty()]));
+  if (toolAtomIds.length === 0)
+    return out;
+  const edgeRows = rowsOf(await transport.request(PULSE_USES_QUERY, { tools: [...toolAtomIds], uses: USES2, limit }));
+  const edges = new Map;
+  const NESTED_CAP = 100;
+  let truncated = edgeRows.length >= limit;
+  for (const row of edgeRows) {
+    const tripleId = readString(row, "term_id");
+    const tool = readString(row, "object_id");
+    const writer = readString(row, "creator_id");
+    const subject = readRecord(row, "subject");
+    const atomId = subject === null ? null : readString(subject, "term_id");
+    if (tripleId === null || tool === null || writer === null || subject === null || atomId === null)
+      continue;
+    const edgeHolders = (Array.isArray(row["positions"]) ? row["positions"].filter(isRecord) : []).map((holder) => readString(holder, "account_id")).filter((account) => account !== null);
+    const declaration = (wallet, declaredAt) => ({
+      tripleId,
+      agent: { atomId, label: readString(subject, "label") },
+      writer: wallet,
+      holders: edgeHolders,
+      declaredAt,
+      domains: [],
+      outcomes: [],
+      lines: []
+    });
+    const declarers = new Map([[key(writer), declaration(writer, readString(row, "created_at"))]]);
+    const holders = Array.isArray(row["positions"]) ? row["positions"].filter(isRecord) : [];
+    if (holders.length >= NESTED_CAP)
+      truncated = true;
+    for (const holder of holders) {
+      const wallet = readString(holder, "account_id");
+      if (wallet !== null && !declarers.has(key(wallet)))
+        declarers.set(key(wallet), declaration(wallet, null));
+    }
+    edges.set(key(tripleId), { tool: key(tool), declarers });
+  }
+  const payers = new Set(config.payers.map(key));
+  const writers = [...new Set([...edges.values()].flatMap((edge) => [...edge.declarers.values()].map((row) => row.writer)))];
+  const qualifiers = async (subjects, predicates) => subjects.length === 0 ? [] : rowsOf(await transport.request(PULSE_QUALIFIERS_QUERY, { edges: subjects, predicates, writers, limit }));
+  const pass = (rows) => {
+    truncated = truncated || rows.length >= limit;
+    return rows;
+  };
+  const standing = (row, edge, below) => {
+    const holders = Array.isArray(row["positions"]) ? row["positions"].filter(isRecord) : [];
+    if (holders.length >= NESTED_CAP)
+      truncated = true;
+    const wallets = [readString(row, "creator_id"), ...holders.map((holder) => readString(holder, "account_id"))];
+    return new Set(wallets.filter((wallet) => wallet !== null).map(key).filter((wallet) => edge.declarers.has(wallet) && !payers.has(wallet) && (below === undefined || below.has(wallet))));
+  };
+  const parts = (row) => {
+    const object = readRecord(row, "object");
+    return {
+      termId: readString(row, "term_id"),
+      subjectId: readString(row, "subject_id"),
+      predicateId: readString(row, "predicate_id"),
+      writer: readString(row, "creator_id"),
+      value: object === null ? null : readString(object, "label")
+    };
+  };
+  const isOutcome = (value) => USAGE_OUTCOMES3.some((word) => word === value);
+  const isReason = (value) => USAGE_REASONS3.some((word) => word === value);
+  const contexts = new Map;
+  const outcomes = new Map;
+  const others = (edge, row, predicate) => out.get(edge.tool)?.byOthers.push({
+    tripleId: row.termId ?? "",
+    writer: row.writer ?? "",
+    predicate,
+    value: row.value ?? ""
+  });
+  const onEdge = pass(await qualifiers([...edges.keys()], [config.inContextOf, config.hasOutcome]));
+  for (const row of onEdge) {
+    const line = parts(row);
+    const edge = line.subjectId === null ? undefined : edges.get(key(line.subjectId));
+    if (edge === undefined || line.predicateId === null || line.writer === null || line.value === null || line.termId === null)
+      continue;
+    const isOutcomeLine = key(line.predicateId) === key(config.hasOutcome);
+    if (isOutcomeLine && !isOutcome(line.value))
+      continue;
+    const wallets = standing(row, edge);
+    if (wallets.size === 0) {
+      others(edge, line, isOutcomeLine ? "outcome" : "context");
+    } else if (isOutcomeLine && isOutcome(line.value)) {
+      outcomes.set(key(line.termId), { edge, domain: null, form: "sibling", outcome: line.value, wallets, reasons: new Map, named: new Map });
+    } else {
+      contexts.set(key(line.termId), { edge, domain: line.value, wallets });
+    }
+  }
+  const onContext = pass(await qualifiers([...contexts.keys()], [config.hasOutcome]));
+  for (const row of onContext) {
+    const line = parts(row);
+    const context = line.subjectId === null ? undefined : contexts.get(key(line.subjectId));
+    if (context === undefined || line.value === null || line.termId === null || !isOutcome(line.value))
+      continue;
+    const wallets = standing(row, context.edge, context.wallets);
+    if (wallets.size === 0) {
+      others(context.edge, line, "outcome");
+    } else {
+      outcomes.set(key(line.termId), { edge: context.edge, domain: context.domain, form: "nested", outcome: line.value, wallets, reasons: new Map, named: new Map });
+    }
+  }
+  const reading = (predicate) => predicate !== undefined;
+  const failures = [...outcomes.entries()].filter(([, found]) => found.outcome === "failed").map(([id]) => id);
+  const worked = [...outcomes.entries()].filter(([, found]) => found.outcome === "worked").map(([id]) => id);
+  const withId = config.with;
+  const onOutcome = pass(await qualifiers(reading(withId) ? [...failures, ...worked] : failures, [config.because, ...reading(withId) ? [withId] : []]));
+  const reasonTerms = new Map;
+  const addName = (found, wallets, name) => {
+    for (const wallet of wallets) {
+      const given = found.named.get(wallet) ?? [];
+      if (!given.includes(name))
+        given.push(name);
+      found.named.set(wallet, given);
+    }
+  };
+  for (const row of onOutcome) {
+    const line = parts(row);
+    const found = line.subjectId === null ? undefined : outcomes.get(key(line.subjectId));
+    if (found === undefined || line.value === null || line.predicateId === null)
+      continue;
+    if (withId !== undefined && key(line.predicateId) === key(withId)) {
+      if (found.outcome !== "worked" || needKindOf3(line.value) === null)
+        continue;
+      const wallets2 = standing(row, found.edge, found.wallets);
+      if (wallets2.size === 0)
+        others(found.edge, line, "with");
+      addName(found, wallets2, line.value);
+      continue;
+    }
+    if (found.outcome !== "failed" || !isReason(line.value))
+      continue;
+    const wallets = standing(row, found.edge, found.wallets);
+    if (wallets.size === 0)
+      others(found.edge, line, "reason");
+    for (const wallet of wallets) {
+      const given = found.reasons.get(wallet) ?? [];
+      if (!given.includes(line.value))
+        given.push(line.value);
+      found.reasons.set(wallet, given);
+    }
+    if (line.termId !== null && wallets.size > 0 && needKindOfReason3(line.value) !== null) {
+      reasonTerms.set(key(line.termId), { found, reason: line.value, wallets });
+    }
+  }
+  const needsId = config.needs;
+  if (reading(needsId)) {
+    const onReason = pass(await qualifiers([...reasonTerms.keys()], [needsId]));
+    for (const row of onReason) {
+      const line = parts(row);
+      const reason = line.subjectId === null ? undefined : reasonTerms.get(key(line.subjectId));
+      if (reason === undefined || line.value === null)
+        continue;
+      const kind = needKindOf3(line.value);
+      if (kind === null || kind !== needKindOfReason3(reason.reason))
+        continue;
+      const wallets = standing(row, reason.found.edge, reason.wallets);
+      if (wallets.size === 0)
+        others(reason.found.edge, line, "needs");
+      addName(reason.found, wallets, line.value);
+    }
+  }
+  for (const { edge, domain, wallets } of contexts.values()) {
+    for (const wallet of wallets)
+      edge.declarers.get(wallet)?.domains.push(domain);
+  }
+  const ordered = [...outcomes.values()].sort((a, b) => Number(b.form === "nested") - Number(a.form === "nested"));
+  for (const found of ordered) {
+    for (const wallet of found.wallets) {
+      const declaration = found.edge.declarers.get(wallet);
+      if (declaration === undefined)
+        continue;
+      const reasons = found.reasons.get(wallet) ?? [];
+      const named = found.named.get(wallet) ?? [];
+      const field = found.outcome === "failed" ? "needs" : "with";
+      if (!declaration.outcomes.includes(found.outcome))
+        declaration.outcomes.push(found.outcome);
+      const domains = found.form === "nested" ? [found.domain] : declaration.domains.length > 0 ? declaration.domains : [null];
+      for (const domain of domains) {
+        const same = declaration.lines.find((line) => line.domain === domain && line.outcome === found.outcome);
+        if (same === undefined) {
+          declaration.lines.push({ domain, outcome: found.outcome, reasons, form: found.form, ...named.length === 0 ? {} : { [field]: named } });
+        } else {
+          same.reasons = [...new Set([...same.reasons, ...reasons])];
+          if (named.length > 0)
+            same[field] = [...new Set([...same[field] ?? [], ...named])];
+        }
+      }
+    }
+  }
+  for (const { tool, declarers } of edges.values()) {
+    const rows = [...declarers.values()].filter((row, at) => at === 0 || row.domains.length > 0 || row.lines.length > 0);
+    out.get(tool)?.declarations.push(...rows);
+  }
+  for (const usage of out.values()) {
+    usage.truncated = truncated;
+    usage.agents = new Set(usage.declarations.map((declaration) => key(declaration.agent.atomId))).size;
+    usage.wallets = new Set(usage.declarations.map((declaration) => key(declaration.writer)).filter((wallet) => !payers.has(wallet))).size;
+    for (const declaration of usage.declarations) {
+      for (const domain of declaration.domains)
+        usage.byDomain[domain] = (usage.byDomain[domain] ?? 0) + 1;
+      for (const outcome of declaration.outcomes)
+        usage.byOutcome[outcome] = (usage.byOutcome[outcome] ?? 0) + 1;
+      for (const reason of new Set(declaration.lines.flatMap((line) => line.reasons))) {
+        usage.byReason[reason] = (usage.byReason[reason] ?? 0) + 1;
+      }
+      const byNeed = usage.byNeed ??= {};
+      for (const name of new Set(declaration.lines.flatMap((line) => line.needs ?? [])))
+        byNeed[name] = (byNeed[name] ?? 0) + 1;
+      const byWith = usage.byWith ??= {};
+      for (const name of new Set(declaration.lines.flatMap((line) => line.with ?? [])))
+        byWith[name] = (byWith[name] ?? 0) + 1;
+    }
+  }
+  return out;
+}
+
+// ../../packages/protocol/src/failure-lines.ts
+function skillUrlOf2(source, skill) {
+  if (!/^[\w.-]+\/[\w.-]+$/.test(source))
+    return null;
+  const name = skill.split(":").pop() ?? skill;
+  const url = `https://github.com/${source}?skill=${encodeURIComponent(name)}`;
+  return skillRefFromUrl3(url) === null ? null : url;
+}
+function skillAtom(url, sha256) {
+  const thing = toolThing2(url);
+  return { uri: thingUriOfDigest(sha256(thingDocument2(thing))), label: thing.name };
+}
+async function readSkillLines(atomUri, config) {
+  const words = [HAS_OUTCOME_LABEL3, BECAUSE_LABEL3, NEEDS_LABEL3, WITH_LABEL3];
+  const atoms = await findToolAtoms2([atomUri, ...words], config);
+  const tool = atoms.get(atomUri);
+  const hasOutcome = atoms.get(HAS_OUTCOME_LABEL3);
+  const because = atoms.get(BECAUSE_LABEL3);
+  if (tool === undefined || hasOutcome === undefined || because === undefined)
+    return null;
+  const needs = atoms.get(NEEDS_LABEL3);
+  const withId = atoms.get(WITH_LABEL3);
+  const usage = await getPulseUsage2([tool], {
+    ...config,
+    hasOutcome,
+    because,
+    ...needs === undefined ? {} : { needs },
+    ...withId === undefined ? {} : { with: withId }
+  });
+  return usage.get(tool.toLowerCase()) ?? null;
+}
+function counted(counts) {
+  return [...counts.entries()].sort(([a, x], [b, y]) => y.size - x.size || a.localeCompare(b)).map(([name, wallets]) => `${name} (${wallets.size})`).join(", ");
+}
+var plural2 = (n) => `${n} wallet${n === 1 ? "" : "s"}`;
+function linesSentence(usage, label, where) {
+  if (usage === null)
+    return null;
+  const parts = [];
+  for (const outcome of ["failed", "worked"]) {
+    const wallets = new Set;
+    const because = new Map;
+    const named = new Map;
+    const add = (to, name, wallet) => to.set(name, (to.get(name) ?? new Set).add(wallet));
+    for (const row of usage.declarations) {
+      const wallet = row.writer.toLowerCase();
+      for (const line of row.lines) {
+        if (line.outcome !== outcome)
+          continue;
+        wallets.add(wallet);
+        for (const reason of line.reasons)
+          add(because, reason, wallet);
+        for (const name of (outcome === "failed" ? line.needs : line.with) ?? [])
+          add(named, name, wallet);
+      }
+    }
+    if (wallets.size === 0)
+      continue;
+    const detail = [
+      ...because.size === 0 ? [] : [`because ${counted(because)}`],
+      ...named.size === 0 ? [] : [`${outcome === "failed" ? "needs" : "with"} ${counted(named)}`]
+    ];
+    parts.push([`${usage.truncated ? "at least " : ""}${plural2(wallets.size)} wrote ${outcome}`, ...detail].join(" · "));
+  }
+  if (parts.length === 0)
+    return null;
+  return `ARP, lines agents wrote on ${label} (${where}): ${parts.join("; ")}.`;
+}
+// ../../data/deployments/13579.json
+var _13579_default = {
+  chain: {
+    name: "Intuition Testnet",
+    chainId: 13579,
+    chainIdHex: "0x350b",
+    nativeCurrency: {
+      name: "Test Trust",
+      symbol: "tTRUST",
+      decimals: 18
+    },
+    rpcUrl: "https://testnet.rpc.intuition.systems/http",
+    explorerUrl: "https://testnet.explorer.intuition.systems",
+    graphqlUrl: "https://testnet.intuition.sh/v1/graphql",
+    _source: [
+      "Intuition skill: .claude/skills/intuition/SKILL.md (table at line 149, Viem chain config at line 184)",
+      "MetaMask deployments: ~/Project/metamask-smart-accounts-kit/packages/delegation-deployments/src/index.ts (CHAIN_ID.intuitionTestnet = 0x350b)"
+    ]
+  },
+  intuition: {
+    _source: ".claude/skills/intuition/SKILL.md line 152",
+    multiVault: "0x2Ece8D4dEdcB9918A398528f3fa4688b1d2CAB91"
+  },
+  metamaskDelegationFramework: {
+    version: "1.3.0",
+    _source: "MetaMask deployments package (~/Project/metamask-smart-accounts-kit/packages/delegation-deployments/src/contractAddresses.ts → DEPLOYMENTS_1_3_0). All addresses are deterministic CREATE2 — same value on every chain MetaMask supports. Intuition Testnet (chainId 0x350b) is explicitly mapped to DEPLOYMENTS_1_3_0 at packages/delegation-deployments/src/index.ts line 148.",
+    delegationManager: "0xdb9B1e94B5b69Df7e401DDbedE43491141047dB3",
+    entryPoint: "0x0000000071727De22E5E9d8BAf0edAc6f37da032",
+    simpleFactory: "0x69Aa2f9fe1572F1B640E1bbc512f5c3a734fc77c",
+    implementations: {
+      hybridDeleGator: "0x48dBe696A4D990079e039489bA2053B36E8FFEC4",
+      multiSigDeleGator: "0x56a9EdB16a0105eb5a4C54f4C062e2868844f3A7",
+      eip7702StatelessDeleGator: "0x63c0c19a282a1B52b07dD5a65b58948A07DAE32B"
+    },
+    enforcers: {
+      _note: "Standard caveat enforcers. ARP composes its custom DomainScopeEnforcer + TrustStakeCapEnforcer alongside a subset of these.",
+      allowedCalldata: "0xc2b0d624c1c4319760C96503BA27C347F3260f55",
+      allowedMethods: "0x2c21fD0Cb9DC8445CB3fb0DC5E7Bb0Aca01842B5",
+      allowedTargets: "0x7F20f61b1f09b08D970938F6fa563634d65c4EeB",
+      approvalRevocation: "0xe264F1f09A19505a1ca1a86D5b01E8bFdb64324A",
+      blockNumber: "0x5d9818dF0AE3f66e9c3D0c5029DAF99d1823ca6c",
+      deployed: "0x24ff2AA430D53a8CD6788018E902E098083dcCd2",
+      erc20BalanceChange: "0xcdF6aB796408598Cea671d79506d7D48E97a5437",
+      erc20TransferAmount: "0xf100b0819427117EcF76Ed94B358B1A5b5C6D2Fc",
+      erc20PeriodTransfer: "0x474e3Ae7E169e940607cC624Da8A15Eb120139aB",
+      erc20Streaming: "0x56c97aE02f233B29fa03502Ecc0457266d9be00e",
+      erc721BalanceChange: "0x8aFdf96eDBbe7e1eD3f5Cd89C7E084841e12A09e",
+      erc721Transfer: "0x3790e6B7233f779b09DA74C72b6e94813925b9aF",
+      erc1155BalanceChange: "0x63c322732695cAFbbD488Fc6937A0A7B66fC001A",
+      exactCalldataBatch: "0x982FD5C86BBF425d7d1451f974192d4525113DfD",
+      exactCalldata: "0x99F2e9bF15ce5eC84685604836F71aB835DBBdED",
+      exactExecutionBatch: "0x1e141e455d08721Dd5BCDA1BaA6Ea5633Afd5017",
+      exactExecution: "0x146713078D39eCC1F5338309c28405ccf85Abfbb",
+      id: "0xC8B5D93463c893401094cc70e66A206fb5987997",
+      logicalOrWrapper: "0xE1302607a3251AF54c3a6e69318d6aa07F5eB46c",
+      limitedCalls: "0x04658B29F6b82ed55274221a06Fc97D318E25416",
+      nativeBalanceChange: "0xbD7B277507723490Cd50b12EaaFe87C616be6880",
+      argsEqualityCheck: "0x44B8C6ae3C304213c3e298495e12497Ed3E56E41",
+      nativeTokenPayment: "0x4803a326ddED6dDBc60e659e5ed12d85c7582811",
+      nativeTokenTransferAmount: "0xF71af580b9c3078fbc2BBF16FbB8EEd82b330320",
+      nativeTokenStreaming: "0xD10b97905a320b13a0608f7E9cC506b56747df19",
+      nativeTokenPeriodTransfer: "0x9BC0FAf4Aca5AE429F4c06aEEaC517520CB16BD9",
+      nonce: "0xDE4f2FAC4B3D87A1d9953Ca5FC09FCa7F366254f",
+      ownershipTransfer: "0x7EEf9734E7092032B5C56310Eb9BbD1f4A524681",
+      redeemer: "0xE144b0b2618071B4E56f746313528a669c7E65c5",
+      specificActionERC20TransferBatch: "0x6649b61c873F6F9686A1E1ae9ee98aC380c7bA13",
+      timestamp: "0x1046bb45C8d673d4ea75321280DB34899413c069",
+      valueLte: "0x92Bf12322527cAA612fd31a0e810472BBB106A8F",
+      multiTokenPeriod: "0xFB2f1a9BD76d3701B730E5d69C3219D42D80eBb7"
+    }
+  },
+  arp: {
+    endpointBinding: {
+      _note: "Task 39c, ADR 0094/0095. The two predicates the endpoint binding is written with: (agent identity atom) has endpoint (endpoint atom), and (pinned binding document) in context of (that triple). Both READ FROM THE GRAPH, neither minted by ARP: scripts/measure/endpoint-predicate.ts, run 2026-10-30, found `has endpoint` (Thing, 6 triples) and `in context of` (Thing, 29 triples) on testnet 13579, and none of `reachable at`, `is reachable at`, `served at`, `hosted at`. `in context of` differs from mainnet's 0x892054b0… (scripts/measure/nested-triples.ts), so never copy it across chains.",
+      measuredAt: "2026-10-30",
+      hasEndpoint: "0xa4488686eb3702e4e972a971ea008d6832a7f935b942366f1091566d1993a5f1",
+      inContextOf: "0x61a88b9c372c0d164d2caf66947b67ed0fcb4c457178a271b6b3dc39fb1f8862"
+    },
+    _note: "Written by scripts/deploy/deploy.sh --broadcast. Addresses are EOA-nonce-derived (CREATE). deploy.sh skips a contract whose address here already has code unless --force, and on a redeploy keeps the replaced address under <name>V<n>. ModuleRegistry and DomainScopeEnforcer were removed from src/ in 59ac83a; their entries are the record.",
+    deployer: "0xE596096F4176b682E300d73963e7B04B383C1AA1",
+    deployedAt: "2026-06-01T13:30:00Z",
+    moduleRegistry: "0xc9a2f66775828017e984E8be077fA2d17e0A41F4",
+    moduleRegistryV1: "0x777C28eCb4688D647B535098d11fB87A9746334f",
+    _moduleRegistryNote: "Redeployed 2026-06-03 with schemaURI uniqueness (ADR 0013). V1 address kept for historical reference; orphan (UI reads V2).",
+    domainScopeEnforcer: "0x516B82E29e3Ca46Ca810FC2EEf348932b198f7f9",
+    trustStakeCapEnforcer: "0x19e07280F1483Aa9aCDb1a7F8C61bfd20c7879Bc",
+    identityRegistry: "0xC165A2AD2E540A4069E02834009161E2b4490d5A",
+    sponsor: "0x77939A4c276f3E549276CEB3A09CDa1fF58b7ec9",
+    sponsorV1: "0xE596096F4176b682E300d73963e7B04B383C1AA1",
+    _sponsorNote: "The address ARP's sponsoring service sends from on 13579 (apps/sponsor, ADR 0092, 0094). Since 2026-11-09 a key of its own, for the hosted service (task 78, ADR 0149): sponsorV1 is the deployer's key, which sent for it before and also holds mainnet funds. Not a contract: an EOA whose key is a file mounted into the service's container, never read by an agent. Before 2026-11-09: The address ARP's sponsoring service sends from on 13579 (apps/sponsor, ADR 0092, 0094): its GET / answered this on 2026-11-02. /tools counts every deposit it sent in Usage: the proof-gated ones, the edge creations it paid for, and, because the address is also ARP's deployer, 3 deposits written before the service existed (scripts/measure/sponsor-deposit-provenance.ts, 2026-11-03). Not a contract: an EOA whose key is in a 0600 file the agent never reads.",
+    deployedTxs: {
+      moduleRegistry: "0x6ee4f4d76dd55c69ea434f8c3c204cb6149df1c588910b1792db1650681fbc2e",
+      moduleRegistryV1: "0x02abe9a0c598df131995bc40927a4e0be0b86e7bdbfe747cfdf37fb19b069634",
+      domainScopeEnforcer: "0x6ca2c4e680b45eaa37147c2458bd38d756e6ffabb560755657cc9caf893df272",
+      trustStakeCapEnforcer: "0x8dd54ea3eeba14645531c501c3f1b8ec95e6a95788825e00fbe239ff32263159",
+      identityRegistry: "0x60b1f6ff7dbdcbca6864794eb9c5cf1e3ba94770daac6d2e77a4df00d726900c",
+      trustStakeCapEnforcerV1: "0x80e36828202a9c7606b92146076ed123155a94cfda3a2ab53013bd27be71764f"
+    },
+    explorerLinks: {
+      moduleRegistry: "https://testnet.explorer.intuition.systems/address/0xc9a2f66775828017e984E8be077fA2d17e0A41F4",
+      moduleRegistryV1: "https://testnet.explorer.intuition.systems/address/0x777C28eCb4688D647B535098d11fB87A9746334f",
+      domainScopeEnforcer: "https://testnet.explorer.intuition.systems/address/0x516B82E29e3Ca46Ca810FC2EEf348932b198f7f9",
+      trustStakeCapEnforcer: "https://testnet.explorer.intuition.systems/address/0x19e07280F1483Aa9aCDb1a7F8C61bfd20c7879Bc",
+      identityRegistry: "https://testnet.explorer.intuition.systems/address/0xC165A2AD2E540A4069E02834009161E2b4490d5A",
+      trustStakeCapEnforcerV1: "https://testnet.explorer.intuition.systems/address/0x7BB56819E9a413B8B4668C5cAF5C494c41dC0F8E"
+    },
+    modules: [
+      {
+        id: "1",
+        name: "Solidity Audit",
+        domain: "solidity-audit",
+        schemaURI: "ipfs://bafkreiembmtprqzufrdkvig74lskv6iazuatblm4hwmtxkf5a3v67mkdkm",
+        creator: "0xE596096F4176b682E300d73963e7B04B383C1AA1",
+        tx: "0xbd97979b197e4423293180e36d69f72e8eb07d2bb11bdfb2940ec918d6f6af3c"
+      }
+    ],
+    intuitionAtoms: [
+      {
+        atomId: "0xde8890fca8dde322ca57bce0e20dfeb23be67332f9101305ebe72facdee032dc",
+        uri: "ipfs://bafkreiembmtprqzufrdkvig74lskv6iazuatblm4hwmtxkf5a3v67mkdkm",
+        tx: "0x97db7629c6c973d43c0c62325754d750ee262853bdbd87496b4b82c1ea3d0ac7"
+      }
+    ],
+    rigAgents: {
+      _note: "Task 24. ERC-8004 agents on ARP's own testnet registry, minted 2026-10-19 on chain 13579. SIGNED BY THE AGENT SESSION using the operator's PRIVATE_KEY from .env, on the user's explicit authorisation — not by a human at a keyboard. See ADR 0037. security.md scopes its automation prohibition to mainnet, so this is permitted and was previously unwritten. Agents 1 and 2 were left with an EMPTY tokenURI deliberately: no-token-uri is 23% of the real cohort (32/140, scripts/measure/cohort-endpoints.ts), so they exercise a branch the three new agents cannot. Permanent: the ids are taken and setAgentURI leaves the superseded URI in the mint transaction's calldata.",
+      registry: "0xC165A2AD2E540A4069E02834009161E2b4490d5A",
+      agents: {
+        "3": {
+          slug: "rig-confirmed",
+          exercises: "confirmed/serves-declared-protocol + tools/list recovery",
+          endpoint: "http://localhost:3002/mcp",
+          _endpointNote: "ARP's own MCP server, local by default. ARP_MCP_ENDPOINT regenerates the file and setAgentURI can update the agent — but that escape hatch is a rehearsal prop: portability.md rule 4 says any step that works because ARP can write to the registry is not a step, and on mainnet ARP owns no registry. The real fix is a public endpoint.",
+          tx: "0xce4fee69c692ddf57d4537b23479d78384a9e2d06b733b532e419055ed50a086",
+          documentSha256_16: "478ac85cd6b2357f",
+          _documentShaNote: "sha256 of the data: URI, first 16 hex. Lets repo-vs-chain drift be detected: the fixtures in scripts/seed/rig-registration-files.ts are the source of truth for three permanent mints, and an edit to FILES would otherwise diverge silently. Checked by scripts/verify/verify-rig-claims.ts against a gate report. ONLY REPRODUCIBLE WITH ARP_MCP_ENDPOINT UNSET: OWN_ENDPOINT reads the environment and the document is hashed after substitution, so anyone with it exported sees a false DRIFT."
+        },
+        "4": {
+          slug: "rig-dead",
+          exercises: "confirmed/dead",
+          endpoint: "http://192.0.2.1/mcp",
+          _endpointNote: "TEST-NET-1, RFC 5737 — reserved and never routable, so it always times out. A .invalid hostname was tried first and gave inconclusive/transport-error, because a DNS failure is indistinguishable from our own network being down.",
+          tx: "0x87bf71cf0ce55db7f82ca617b8601ee87f870e67018b1097a311ed2d61cc7378",
+          uriUpdatedTx: "0x8f5950a61fca19b6a3f69b26f6c4159e29833689ed4b1623d8610206bfe87100",
+          documentSha256_16: "6db8b769dd551e98"
+        },
+        "5": {
+          slug: "rig-unprobeable",
+          exercises: "not-checkable ×2 — templated and non-http-identifier. Never a defect.",
+          tx: "0x859aff609fa10dba635ffd12d0b60b598c7c3174238f61962edb4b5eee1a0efd",
+          documentSha256_16: "f61a9d7f6637958e"
+        }
+      },
+      gate: {
+        command: "bun run scripts/verify/verify-cohort.ts --target intuition-testnet --ids 1,2,3,4,5 --tools --out /tmp/gate.json && bun run scripts/verify/verify-rig-claims.ts /tmp/gate.json",
+        ranAt: "2026-10-19",
+        result: "5 agents: 2 with no tokenURI (40%), 3 read. 4 endpoints: 2 confirmed (serves-declared-protocol, dead), 2 not-checkable (templated, non-http-identifier), 0 inconclusive, 1 capability list recovered (arp_tool_reputation).",
+        _requires: "apps/mcp/src/server.ts listening on :3002, or agent 3 reports inconclusive. The confirmed branch is therefore reproducible only on a host running that server — a known limit of this gate until ARP has a public endpoint.",
+        _widerThanSpecced: "--ids 1,2,3,4,5 rather than 3,4,5. Including the two empty-tokenURI agents exercises a fifth branch for two extra requests.",
+        _claimsCheck: "verify-rig-claims.ts asserts each fixture's `exercises` prose against the outcome the probe observed, and the document hashes against this file. The unit suite CANNOT close that class: confirmed and inconclusive are both probeable before a request is sent, and the mutation that actually shipped lived exactly there."
+      }
+    },
+    trustStakeCapEnforcerV1: "0x7BB56819E9a413B8B4668C5cAF5C494c41dC0F8E",
+    delegationRegister: {
+      _note: "Task 37c, ADR 0092 item 7. The two predicates a published delegation is written with: (delegator CAIP-10) delegate to (delegate CAIP-10), and (pinned document) in context of (that triple). Both READ FROM THE GRAPH, neither minted by ARP: on testnet 13579 on 2026-11-03 exactly one atom carries each label, a Thing, `delegate to` in 10 triples and `in context of` in 29. `in context of` is the one endpointBinding uses. Never copy either across chains.",
+      measuredAt: "2026-11-03",
+      delegateTo: "0xb56980d42a3b03455bf41ea20fe04ae223fca0b9e688994dc661414e81e6433b",
+      inContextOf: "0x61a88b9c372c0d164d2caf66947b67ed0fcb4c457178a271b6b3dc39fb1f8862"
+    }
+  }
+};
+
+// ../hook/src/skill-failure.ts
+import { existsSync, mkdirSync as mkdirSync2, readFileSync as readFileSync2 } from "node:fs";
+import { homedir as homedir2 } from "node:os";
+import { dirname as dirname2, join as join3 } from "node:path";
+
+// ../../packages/protocol/src/declare/label.ts
+function normaliseCategoryLabel2(raw) {
+  return raw.normalize("NFKC").normalize("NFD").replace(/[̀-ͯ]/g, "").toLowerCase().trim().replace(/\s+/g, " ");
+}
+
+// ../../packages/protocol/src/domain-words.ts
+function tokens(text) {
+  return normaliseCategoryLabel2(text).split(/[^a-z0-9]+/).filter((t) => t.length > 1);
+}
+function domainInText2(text, words) {
+  const padded = ` ${tokens(text).join(" ")} `;
+  const hit = (phrase) => padded.includes(` ${tokens(phrase).join(" ")} `);
+  const label = [...words.labels].filter((l) => tokens(l).length > 0 && hit(l)).sort((a, b) => b.length - a.length)[0];
+  if (label !== undefined)
+    return label;
+  const alias = Object.keys(words.aliases).filter((a) => words.labels.includes(words.aliases[a]) && hit(a)).sort((a, b) => b.length - a.length)[0];
+  return alias === undefined ? null : words.aliases[alias];
+}
+// ../../data/catalogue/domain-aliases.json
+var domain_aliases_default = {
+  _note: "Words a person or an agent uses, mapped to a node of ARP's domain tree (OASF 1.1.0 names, plus ARP's own `design`), task 84. Read locally by the MCP server, the CLI and the hook: the words never leave the machine (the user, 2026-11-09), except for a caller who chooses ARP's hosted read-only endpoint, which receives its words and keeps none (ADR 0161). Every target is checked against data/catalogue/oasf-1.1.0-skills.json by packages/protocol/test/domain-words.test.ts. ARP's old category words map to the one node that held them; the split ones (method, quality) have no alias, and a refusal suggests nodes instead.",
+  oasfVersion: "1.1.0",
+  aliases: {
+    ads: "advertising",
+    advertising: "advertising",
+    agent: "agent application building",
+    agents: "agent orchestration",
+    android: "mobile development",
+    animation: "web animation",
+    api: "api development",
+    audio: "audio & speech processing",
+    automation: "tool use and automation",
+    backend: "backend development",
+    blockchain: "blockchain and web3",
+    brand: "branding",
+    browser: "browser automation",
+    captions: "video editing",
+    chart: "chart generation",
+    charts: "chart generation",
+    chat: "team messaging",
+    collaboration: "collaboration and communication",
+    commit: "commit authoring",
+    copywriting: "copywriting",
+    crawl: "web crawling",
+    cro: "conversion optimization",
+    dashboard: "dashboard building",
+    database: "databases",
+    debug: "code debugging",
+    debugging: "code debugging",
+    deck: "presentation authoring",
+    deploy: "deployment and release management",
+    development: "software engineering",
+    docker: "containerization",
+    docs: "technical writing",
+    document: "document authoring",
+    documentation: "project documentation",
+    docx: "document authoring",
+    editing: "video editing",
+    email: "email marketing",
+    excel: "spreadsheet authoring",
+    frontend: "frontend development",
+    git: "git workflow",
+    illustration: "image generation",
+    image: "computer vision",
+    infrastructure: "devops and cloud infrastructure",
+    ios: "mobile development",
+    issues: "issue tracking",
+    kubernetes: "container orchestration",
+    landing: "conversion optimization",
+    layout: "design",
+    logo: "branding",
+    logs: "logging and log analysis",
+    marketing: "content, writing and marketing",
+    mcp: "agent tool building",
+    meeting: "meeting intelligence",
+    meetings: "meeting intelligence",
+    mobile: "mobile development",
+    monitoring: "monitoring and alerting",
+    montage: "video editing",
+    music: "text to audio",
+    newsletter: "email marketing",
+    notes: "note-taking",
+    payment: "payments integration",
+    pdf: "pdf manipulation",
+    pentest: "penetration testing",
+    postgres: "relational database operations",
+    powerpoint: "presentation authoring",
+    pptx: "presentation authoring",
+    pr: "pull request management",
+    prd: "requirements and prd authoring",
+    presentation: "presentation authoring",
+    prompt: "prompt design",
+    rag: "retrieval-augmented generation",
+    research: "research and scholarship",
+    review: "code review",
+    scrape: "web scraping",
+    scraping: "web data acquisition",
+    search: "web search research",
+    security: "cybersecurity",
+    seo: "search engine optimization",
+    slide: "presentation authoring",
+    slides: "presentation authoring",
+    "smart contract": "smart contract development",
+    social: "social media marketing",
+    solidity: "smart contract development",
+    spec: "requirements and prd authoring",
+    spreadsheet: "spreadsheet authoring",
+    sql: "relational database operations",
+    subtitles: "video editing",
+    summarize: "text summarization",
+    summary: "text summarization",
+    tdd: "software testing",
+    test: "software testing",
+    testing: "software testing",
+    tests: "software testing",
+    thumbnail: "image generation",
+    tickets: "issue tracking",
+    trading: "defi and crypto trading",
+    transcription: "automatic speech recognition",
+    translate: "translation",
+    tts: "text to speech",
+    typography: "design",
+    ui: "design",
+    ux: "design",
+    voice: "text to speech",
+    wallet: "crypto wallet operations",
+    website: "web development",
+    writing: "content writing",
+    xlsx: "spreadsheet authoring"
+  }
+};
+// ../../data/catalogue/skill-nodes-13579.json
+var skill_nodes_13579_default = {
+  _note: "For the hook's failure record, which reads no network (task 84): the nodes of ARP's testnet tree as written by task 79 (data/catalogue/oasf-tree-13579.json), and the node each skill name is filed at, when the name is filed at one node only. Built by scripts/build/build-skill-nodes.ts from data/catalogue/oasf-skill-tags-proposal-13579.json, rows not reviewed out.",
+  chainId: 13579,
+  oasfVersion: "1.1.0",
+  nodes: [
+    "ad copywriting",
+    "ad creative production",
+    "advertising",
+    "agent application building",
+    "agent evaluation",
+    "agent orchestration",
+    "agent tool building",
+    "agentic workflow orchestration",
+    "ai agent development",
+    "ai safety and guardrails",
+    "ai/ml engineering",
+    "answer engine optimization",
+    "api development",
+    "api schema understanding",
+    "app store release",
+    "architecture patterns",
+    "architecture review",
+    "article and blog writing",
+    "audio & speech processing",
+    "audio generation",
+    "audio to audio",
+    "authentication implementation",
+    "authorization and access control",
+    "benchmark execution",
+    "branding",
+    "browser automation",
+    "build, packaging and dev tooling",
+    "business and professional",
+    "business and professional writing",
+    "business operations",
+    "business strategy",
+    "calendar and scheduling",
+    "campaign planning",
+    "ci/cd configuration",
+    "ci/cd pipelines",
+    "cli and terminal app development",
+    "cloud architecture design",
+    "cloud cost optimization",
+    "cloud networking",
+    "cloud platform operations",
+    "cloud resource management",
+    "code comprehension",
+    "code debugging",
+    "code generation",
+    "code migration and modernization",
+    "code quality and maintenance",
+    "code refactoring and optimization",
+    "code review",
+    "collaboration and communication",
+    "commit authoring",
+    "competitive analysis",
+    "competitive intelligence",
+    "compliance assessment",
+    "computer vision",
+    "configuration management",
+    "container orchestration",
+    "container workload management",
+    "containerization",
+    "content strategy",
+    "content writing",
+    "content, writing and marketing",
+    "context window management",
+    "controllable image generation",
+    "conversion optimization",
+    "copy editing",
+    "copywriting",
+    "crm management",
+    "cross-platform mobile development",
+    "customer support and success",
+    "cybersecurity",
+    "data analytics",
+    "data engineering and analytics",
+    "data modeling",
+    "data modeling and design",
+    "data pipelines",
+    "data visualization",
+    "data warehousing",
+    "databases",
+    "deep learning training",
+    "deep research",
+    "dependency management",
+    "deployment and release management",
+    "deployment orchestration",
+    "design",
+    "dev environment setup",
+    "devops and cloud infrastructure",
+    "document authoring",
+    "document collaboration",
+    "document or database question answering",
+    "document processing",
+    "editing and proofreading",
+    "email campaign creation",
+    "email management",
+    "email marketing",
+    "employee lifecycle",
+    "end-to-end and ui testing",
+    "error handling implementation",
+    "experiment tracking",
+    "face swapping",
+    "frontend development",
+    "full-stack development",
+    "game development",
+    "git workflow",
+    "governance, compliance and ethics",
+    "graph database operations",
+    "growth strategy",
+    "human resources",
+    "iac authoring",
+    "identity administration",
+    "identity and access management",
+    "image editing",
+    "image generation",
+    "image inpainting",
+    "image outpainting",
+    "image to image",
+    "image to video",
+    "incident response",
+    "inference and serving",
+    "infrastructure as code",
+    "instructed image editing",
+    "interactive visualization",
+    "issue tracking",
+    "knowledge base management",
+    "knowledge management",
+    "landing page and sales copy",
+    "language generation",
+    "language processing",
+    "lead generation",
+    "llm evaluation",
+    "llm observability",
+    "llm tracing and instrumentation",
+    "logging and log analysis",
+    "logical reasoning",
+    "market and competitive intelligence",
+    "market research",
+    "marketing copywriting",
+    "marketing strategy",
+    "meeting intelligence",
+    "ml experimentation and lifecycle",
+    "mobile development",
+    "model and agent evaluation",
+    "model fine-tuning",
+    "model serving",
+    "model training and fine-tuning",
+    "monitoring and alerting",
+    "multi-agent coordination",
+    "native mobile development",
+    "network configuration",
+    "nosql database operations",
+    "note-taking",
+    "observability",
+    "observability instrumentation",
+    "office productivity",
+    "orm and data access",
+    "paid ads management",
+    "paper analysis",
+    "pdf manipulation",
+    "performance monitoring",
+    "postmortem authoring",
+    "presentation authoring",
+    "problem solving",
+    "process operations",
+    "product analytics",
+    "product management",
+    "product marketing and gtm",
+    "product strategy",
+    "production llm monitoring",
+    "project and task management",
+    "project scaffolding",
+    "prompt design",
+    "prompt engineering",
+    "prompt optimization",
+    "pull request management",
+    "quality evaluation",
+    "reasoning and planning",
+    "relational database operations",
+    "release management",
+    "requirements and prd authoring",
+    "research and scholarship",
+    "research ideation",
+    "research, knowledge and productivity",
+    "responsible ai governance",
+    "retention management",
+    "retrieval-augmented generation",
+    "roadmapping and prioritization",
+    "role assignment",
+    "root-cause debugging",
+    "runtime guardrails",
+    "sales and crm",
+    "sales enablement",
+    "sales outreach",
+    "script integration",
+    "search engine optimization",
+    "security hardening",
+    "security operations",
+    "seo auditing",
+    "seo content optimization",
+    "serverless and edge computing",
+    "serverless function development",
+    "site reliability engineering",
+    "social content creation",
+    "social media management",
+    "social media marketing",
+    "software architecture and design",
+    "software engineering",
+    "software testing",
+    "specialized application development",
+    "spreadsheet authoring",
+    "static analysis and linting",
+    "stream processing",
+    "structured data markup",
+    "styling and responsive layout",
+    "support operations",
+    "system design",
+    "talking avatar generation",
+    "task decomposition",
+    "task management",
+    "team messaging",
+    "technical writing",
+    "test strategy and tdd",
+    "text paraphrasing",
+    "text style transfer",
+    "text to audio",
+    "text to code",
+    "text to image",
+    "text to video",
+    "third-party integration",
+    "threat detection",
+    "tool use and automation",
+    "tool use planning",
+    "trace analysis and debugging",
+    "ux writing and microcopy",
+    "version control and collaboration",
+    "video",
+    "video editing",
+    "vulnerability analysis",
+    "vulnerability management",
+    "web accessibility",
+    "web animation",
+    "web development",
+    "web search",
+    "web search research",
+    "workflow automation"
+  ],
+  skills: {
+    "ab-test-setup": "product analytics",
+    "ab-testing": "product analytics",
+    "academy-guide": "support operations",
+    "ace-step": "text to audio",
+    "ad-creative": "ad copywriting",
+    adapt: "styling and responsive layout",
+    ads: "paid ads management",
+    "agent-browser": "browser automation",
+    "agent-tools": "script integration",
+    "agents-sdk": "agent application building",
+    "ai-avatar-video": "talking avatar generation",
+    "ai-image-generation": "text to image",
+    "ai-music": "text to audio",
+    "ai-paper-reproduction": "research and scholarship",
+    "ai-research-explore": "research ideation",
+    "ai-research-reproduction": "research and scholarship",
+    "ai-seo": "answer engine optimization",
+    "ai-video-generation": "text to video",
+    "airunway-aks-setup": "model serving",
+    "algorithmic-art": "image generation",
+    analytics: "product analytics",
+    "analytics-tracking": "product analytics",
+    "analyze-project": "code comprehension",
+    "analyze-test-run": "software testing",
+    animate: "web animation",
+    "animate-expo": "web animation",
+    "animation-performance": "web animation",
+    "animation-vocabulary": "web animation",
+    animejs: "web animation",
+    "app-store-screenshots": "app store release",
+    "appinsights-instrumentation": "observability instrumentation",
+    "apple-design": "design",
+    arrange: "design",
+    "ask-matt": "tool use planning",
+    "ask-sonner": "frontend development",
+    aso: "seo auditing",
+    "aso-audit": "seo auditing",
+    attribution: "product analytics",
+    audit: "web accessibility",
+    "azure-ai": "cloud platform operations",
+    "azure-aigateway": "runtime guardrails",
+    "azure-app-onboard": "deployment orchestration",
+    "azure-app-onboard-prereq": "deployment orchestration",
+    "azure-cloud-migrate": "code migration and modernization",
+    "azure-compliance": "compliance assessment",
+    "azure-compute": "cloud resource management",
+    "azure-cost": "cloud cost optimization",
+    "azure-cost-optimization": "cloud cost optimization",
+    "azure-deploy": "deployment orchestration",
+    "azure-diagnostics": "incident response",
+    "azure-enterprise-infra-planner": "cloud architecture design",
+    "azure-hosted-copilot-sdk": "agent application building",
+    "azure-kubernetes": "container workload management",
+    "azure-kusto": "logging and log analysis",
+    "azure-kusto-graph": "graph database operations",
+    "azure-kusto-irql": "threat detection",
+    "azure-kusto-irql-graph": "interactive visualization",
+    "azure-messaging": "code debugging",
+    "azure-observability": "observability",
+    "azure-postgres": "relational database operations",
+    "azure-prepare": "iac authoring",
+    "azure-quotas": "cloud resource management",
+    "azure-rbac": "authorization and access control",
+    "azure-reliability": "site reliability engineering",
+    "azure-resource-lookup": "cloud resource management",
+    "azure-resource-visualizer": "cloud architecture design",
+    "azure-storage": "cloud resource management",
+    "azure-upgrade": "code migration and modernization",
+    "azure-validate": "ci/cd configuration",
+    basin: "data warehousing",
+    "batch-grill-me": "problem solving",
+    bolder: "design",
+    brainstorming: "reasoning and planning",
+    "brand-guidelines": "branding",
+    "break-ui": "end-to-end and ui testing",
+    "building-ai-agent-on-cloudflare": "agent application building",
+    "building-mcp-server-on-cloudflare": "agent tool building",
+    "canopy-part-title": "video editing",
+    "canvas-design": "image generation",
+    "captions-overlay": "video editing",
+    cavecrew: "role assignment",
+    caveman: "text style transfer",
+    "caveman-commit": "commit authoring",
+    "caveman-compress": "context window management",
+    "caveman-discover": "llm tracing and instrumentation",
+    "caveman-es": "text style transfer",
+    "caveman-evidence-review": "trace analysis and debugging",
+    "caveman-explore": "code comprehension",
+    "caveman-help": "text style transfer",
+    "caveman-learn": "prompt optimization",
+    "caveman-manage": "release management",
+    "caveman-optimize": "llm evaluation",
+    "caveman-review": "code review",
+    "caveman-setup": "llm tracing and instrumentation",
+    "caveman-stats": "production llm monitoring",
+    "cf-cli": "cloud resource management",
+    "changelog-video": "video editing",
+    "character-design-sheet": "controllable image generation",
+    "chief-of-staff": "agentic workflow orchestration",
+    "churn-prevention": "retention management",
+    clarify: "ux writing and microcopy",
+    "claude-api": "third-party integration",
+    "claude-design-hyperframes": "video editing",
+    "claude-handoff": "multi-agent coordination",
+    cloudflare: "cloud architecture design",
+    "cloudflare-email-service": "third-party integration",
+    "cloudflare-one": "network configuration",
+    "cloudflare-one-migrations": "network configuration",
+    "co-marketing": "campaign planning",
+    "code-review": "code review",
+    "code-slice-hero": "video editing",
+    "codebase-design": "architecture patterns",
+    "codex-pet": "image to image",
+    "cold-email": "sales outreach",
+    colorize: "design",
+    "community-marketing": "growth strategy",
+    "competitor-alternatives": "landing page and sales copy",
+    "competitor-profiling": "competitive intelligence",
+    "competitor-teardown": "competitive analysis",
+    competitors: "landing page and sales copy",
+    compress: "context window management",
+    "content-strategy": "content strategy",
+    "contribute-catalog": "video editing",
+    "controlnet-pose": "image to video",
+    "conversion-tracking": "product analytics",
+    "copy-editing": "copy editing",
+    copywriting: "marketing copywriting",
+    critique: "design",
+    cro: "conversion optimization",
+    "css-animations": "web animation",
+    "cuboid-carousel": "video editing",
+    "customer-research": "market research",
+    "cut-the-curve": "video editing",
+    "decision-mapping": "task decomposition",
+    delight: "design",
+    "deploy-to-vercel": "deployment orchestration",
+    "design-an-interface": "system design",
+    "design-mobile-apps": "design",
+    "design-taste-frontend": "design",
+    "design-taste-frontend-v1": "design",
+    "design-taste-frontend-v2-experimental": "design",
+    "developing-genkit-dart": "agent application building",
+    "developing-genkit-go": "agent application building",
+    "developing-genkit-js": "agent application building",
+    "developing-genkit-python": "agent application building",
+    diagnose: "root-cause debugging",
+    "diagnosing-bugs": "root-cause debugging",
+    "diagnosing-superpowers": "trace analysis and debugging",
+    "directory-submissions": "search engine optimization",
+    "discernment-nudge": "responsible ai governance",
+    "discover-azure-skills": "tool use planning",
+    "dispatching-parallel-agents": "multi-agent coordination",
+    distill: "design",
+    "doc-coauthoring": "technical writing",
+    docx: "document authoring",
+    "domain-modeling": "data modeling",
+    "durable-objects": "serverless function development",
+    "edit-article": "editing and proofreading",
+    "elevenlabs-music-generation": "text to audio",
+    "email-sequence": "email campaign creation",
+    emails: "email campaign creation",
+    "embedded-captions": "video editing",
+    "emil-design-eng": "design",
+    "entra-agent-id": "identity administration",
+    "entra-app-registration": "authentication implementation",
+    "env-and-assets-bootstrap": "dev environment setup",
+    events: "campaign planning",
+    "executing-plans": "task management",
+    "explore-code": "deep learning training",
+    "explore-run": "experiment tracking",
+    "extension-to-functions-codebase": "code migration and modernization",
+    extract: "design",
+    "face-swap": "face swapping",
+    "faceless-explainer": "text to video",
+    figma: "video editing",
+    "file-test-bug": "issue tracking",
+    "find-animation-opportunities": "web animation",
+    "find-skills": "tool use planning",
+    "finishing-a-development-branch": "git workflow",
+    "firebase-ai-logic": "third-party integration",
+    "firebase-ai-logic-basics": "third-party integration",
+    "firebase-app-hosting-basics": "deployment orchestration",
+    "firebase-auth-basics": "authentication implementation",
+    "firebase-basics": "dev environment setup",
+    "firebase-crashlytics": "monitoring and alerting",
+    "firebase-data-connect": "relational database operations",
+    "firebase-firestore": "nosql database operations",
+    "firebase-firestore-basics": "nosql database operations",
+    "firebase-firestore-enterprise-native-mode": "nosql database operations",
+    "firebase-firestore-standard": "nosql database operations",
+    "firebase-hosting-basics": "deployment orchestration",
+    "firebase-local-env-setup": "dev environment setup",
+    "firebase-remote-config-basics": "configuration management",
+    "firebase-security-rules-auditor": "vulnerability analysis",
+    "firestore-rules-creation": "authorization and access control",
+    "firestore-security-rules-auditor": "vulnerability analysis",
+    "flux-2-klein": "text to image",
+    "flux-kontext": "instructed image editing",
+    "form-cro": "conversion optimization",
+    "free-tool-strategy": "growth strategy",
+    "free-tools": "growth strategy",
+    "frontend-design": "design",
+    "frost-sequence-camera-orbit": "video editing",
+    "full-output-enforcement": "prompt design",
+    "gateway-workers": "serverless function development",
+    "general-video": "video editing",
+    "git-guardrails-claude-code": "git workflow",
+    "glass-shard-title": "video editing",
+    "google-agents-cli-adk-code": "agent application building",
+    "google-agents-cli-deploy": "deployment orchestration",
+    "google-agents-cli-eval": "agent evaluation",
+    "google-agents-cli-observability": "llm tracing and instrumentation",
+    "google-agents-cli-publish": "release management",
+    "google-agents-cli-scaffold": "project scaffolding",
+    "google-agents-cli-workflow": "agent application building",
+    "gpt-image": "text to image",
+    "gpt-image-2": "text to image",
+    "gpt-image-edit": "instructed image editing",
+    "gpt-taste": "design",
+    "graphic-overlays": "video editing",
+    "grill-me": "problem solving",
+    "grill-with-docs": "problem solving",
+    grilling: "problem solving",
+    gsap: "web animation",
+    handoff: "context window management",
+    happyhorse: "text to video",
+    "happyhorse-1-0": "text to video",
+    harden: "error handling implementation",
+    "higgsfield-brandkit": "branding",
+    "higgsfield-faceless-video": "text to video",
+    "higgsfield-game-generation": "game development",
+    "higgsfield-generate": "text to image",
+    "higgsfield-marketplace-cards": "ad creative production",
+    "higgsfield-product-photoshoot": "text to image",
+    "higgsfield-soul-id": "model fine-tuning",
+    "higgsfield-websites": "full-stack development",
+    "higgsfield-youtube-thumbnail": "text to image",
+    "high-end-visual-design": "design",
+    hyperframes: "video",
+    "hyperframes-animation": "web animation",
+    "hyperframes-audio": "audio to audio",
+    "hyperframes-cli": "video editing",
+    "hyperframes-core": "video editing",
+    "hyperframes-creative": "video editing",
+    "hyperframes-keyframes": "video editing",
+    "hyperframes-media": "audio generation",
+    "hyperframes-read-first": "video",
+    "hyperframes-registry": "video editing",
+    "hyperframes-studio": "video editing",
+    image: "image generation",
+    "image-edit": "instructed image editing",
+    "image-inpainting": "image inpainting",
+    "image-outpainting": "image outpainting",
+    "image-taste-frontend": "design",
+    "image-to-video": "image to video",
+    impeccable: "design",
+    implement: "text to code",
+    "implement-spec": "text to code",
+    "improve-animations": "web animation",
+    "improve-codebase-architecture": "architecture review",
+    "industrial-brutalist-ui": "design",
+    "influencer-marketing": "campaign planning",
+    "infsh-cli": "script integration",
+    "internal-comms": "business and professional writing",
+    "investigate-first": "root-cause debugging",
+    k2: "stream processing",
+    "kling-3-0": "text to video",
+    "landing-page-design": "conversion optimization",
+    "lark-approval": "process operations",
+    "lark-apps": "full-stack development",
+    "lark-attendance": "employee lifecycle",
+    "lark-base": "spreadsheet authoring",
+    "lark-calendar": "calendar and scheduling",
+    "lark-contact": "team messaging",
+    "lark-doc": "document collaboration",
+    "lark-drive": "document collaboration",
+    "lark-event": "workflow automation",
+    "lark-im": "team messaging",
+    "lark-mail": "email management",
+    "lark-markdown": "document collaboration",
+    "lark-meeting": "meeting intelligence",
+    "lark-minutes": "meeting intelligence",
+    "lark-note": "meeting intelligence",
+    "lark-okr": "roadmapping and prioritization",
+    "lark-openapi-explorer": "api schema understanding",
+    "lark-shared": "authentication implementation",
+    "lark-sheets": "spreadsheet authoring",
+    "lark-skill-maker": "agent tool building",
+    "lark-slides": "presentation authoring",
+    "lark-task": "task management",
+    "lark-vc": "meeting intelligence",
+    "lark-vc-agent": "meeting intelligence",
+    "lark-whiteboard": "document collaboration",
+    "lark-whiteboard-cli": "document collaboration",
+    "lark-wiki": "knowledge base management",
+    "lark-workflow-meeting-summary": "meeting intelligence",
+    "lark-workflow-standup-report": "task management",
+    launch: "product marketing and gtm",
+    "launch-strategy": "product marketing and gtm",
+    layout: "design",
+    "lead-magnets": "lead generation",
+    "lean-build": "text to code",
+    "link-building": "search engine optimization",
+    lipsync: "talking avatar generation",
+    "loop-me": "requirements and prd authoring",
+    lottie: "web animation",
+    "markdown-token-optimizer": "context window management",
+    "marketing-council": "marketing strategy",
+    "marketing-ideas": "growth strategy",
+    "marketing-loops": "workflow automation",
+    "marketing-plan": "marketing strategy",
+    "marketing-psychology": "marketing strategy",
+    "mcp-builder": "agent tool building",
+    "media-use": "video",
+    mediabunny: "video",
+    megacave: "text style transfer",
+    "microsoft-foundry": "agent application building",
+    "migrate-to-shoehorn": "code migration and modernization",
+    migration: "code migration and modernization",
+    "minimal-run-and-audit": "benchmark execution",
+    "minimalist-ui": "design",
+    "mobile-native": "styling and responsive layout",
+    "motion-doctrine": "video editing",
+    "motion-graphics": "video editing",
+    "music-to-video": "video editing",
+    "nano-banana-2": "text to image",
+    "nano-banana-edit": "instructed image editing",
+    "nextjs-on-cloudflare": "deployment orchestration",
+    normalize: "design",
+    "obsidian-vault": "note-taking",
+    offers: "product marketing and gtm",
+    onboard: "design",
+    onboarding: "conversion optimization",
+    "onboarding-cro": "conversion optimization",
+    optimize: "code refactoring and optimization",
+    "orbit-card": "video editing",
+    overdrive: "web animation",
+    "oversized-cursor": "video editing",
+    "p-image": "text to image",
+    "p-video": "text to video",
+    "p-video-avatar": "talking avatar generation",
+    "page-cro": "conversion optimization",
+    "paid-ads": "paid ads management",
+    "paper-context-resolver": "paper analysis",
+    "paywall-upgrade-cro": "conversion optimization",
+    paywalls: "conversion optimization",
+    pdf: "pdf manipulation",
+    "pick-ui-library": "frontend development",
+    polish: "design",
+    "popup-cro": "conversion optimization",
+    popups: "conversion optimization",
+    pptx: "presentation authoring",
+    pr: "pull request management",
+    "pr-to-video": "video editing",
+    pricing: "product strategy",
+    "pricing-strategy": "product strategy",
+    "prisma-cli": "orm and data access",
+    "prisma-client-api": "orm and data access",
+    "prisma-compute": "deployment orchestration",
+    "prisma-database-setup": "orm and data access",
+    "prisma-driver-adapter-implementation": "orm and data access",
+    "prisma-mongodb-upgrade": "code migration and modernization",
+    "prisma-orm-setup": "orm and data access",
+    "prisma-postgres": "relational database operations",
+    "prisma-postgres-setup": "relational database operations",
+    "prisma-upgrade-v7": "code migration and modernization",
+    "product-hunt-launch": "product marketing and gtm",
+    "product-launch-video": "ad creative production",
+    "product-marketing": "product marketing and gtm",
+    "product-marketing-context": "product marketing and gtm",
+    "product-photography": "text to image",
+    "programmatic-seo": "seo content optimization",
+    prospecting: "lead generation",
+    "public-relations": "marketing strategy",
+    "python-appservice-deploy": "deployment orchestration",
+    "python-executor": "script integration",
+    qa: "issue tracking",
+    quieter: "design",
+    "receiving-code-review": "code review",
+    "reddit-automation": "social content creation",
+    "redesign-existing-projects": "design",
+    "referral-program": "growth strategy",
+    referrals: "growth strategy",
+    relight: "instructed image editing",
+    "remotion-best-practices": "video",
+    "remotion-captions": "video editing",
+    "remotion-create": "video",
+    "remotion-docs": "document or database question answering",
+    "remotion-interactivity": "video",
+    "remotion-maps": "video",
+    "remotion-markup": "video",
+    "remotion-multimedia": "video",
+    "remotion-render": "video",
+    "remotion-saas": "full-stack development",
+    "remotion-studio": "video",
+    "remotion-to-hyperframes": "code migration and modernization",
+    "remotion-upgrade": "dependency management",
+    "remove-animations": "web animation",
+    "repo-intake-and-plan": "code comprehension",
+    "request-refactor-plan": "code refactoring and optimization",
+    "requesting-code-review": "code review",
+    research: "deep research",
+    "resolving-merge-conflicts": "git workflow",
+    retro: "postmortem authoring",
+    review: "code review",
+    "review-animations": "web animation",
+    revops: "crm management",
+    "run-train": "deep learning training",
+    "runcomfy-cli": "script integration",
+    "safe-debug": "root-cause debugging",
+    "safe-refactor": "code refactoring and optimization",
+    "sales-enablement": "sales enablement",
+    "sandbox-migrate-to-next": "code migration and modernization",
+    "sandbox-next": "containerization",
+    "sandbox-sdk": "containerization",
+    "sandbox-stable": "containerization",
+    "scaffold-exercises": "project scaffolding",
+    schema: "structured data markup",
+    "schema-markup": "structured data markup",
+    "seam-craft": "video editing",
+    seedance: "text to video",
+    "seedance-v2": "text to video",
+    sensei: "agent tool building",
+    "seo-audit": "seo auditing",
+    "setup-matt-pocock-skills": "dev environment setup",
+    "setup-pre-commit": "dev environment setup",
+    "setup-ts-deep-modules": "static analysis and linting",
+    shape: "design",
+    signup: "conversion optimization",
+    "signup-flow-cro": "conversion optimization",
+    "site-architecture": "search engine optimization",
+    "skill-authoring": "agent tool building",
+    "skill-creator": "agent tool building",
+    "slack-gif-creator": "image generation",
+    slideshow: "presentation authoring",
+    sms: "email marketing",
+    social: "social media management",
+    "social-content": "social content creation",
+    "stitch-design-taste": "design",
+    "storyboard-creation": "video",
+    "subagent-driven-development": "agentic workflow orchestration",
+    "surgical-patch": "root-cause debugging",
+    "systematic-debugging": "root-cause debugging",
+    tailwind: "styling and responsive layout",
+    "talking-head-recut": "video editing",
+    tdd: "test strategy and tdd",
+    teach: "research and scholarship",
+    "template-skill": "agent tool building",
+    "test-driven-development": "test strategy and tdd",
+    "theme-factory": "design",
+    three: "web animation",
+    "to-issues": "issue tracking",
+    "to-prd": "requirements and prd authoring",
+    "to-questionnaire": "reasoning and planning",
+    "to-spec": "requirements and prd authoring",
+    "to-tickets": "issue tracking",
+    triage: "issue tracking",
+    "turnstile-spin": "security hardening",
+    "twitter-automation": "social media management",
+    typegpu: "web animation",
+    typeset: "design",
+    "ubiquitous-language": "data modeling",
+    ultracave: "text style transfer",
+    "using-git-worktrees": "git workflow",
+    "using-superpowers": "tool use planning",
+    "vercel-cli-with-tokens": "deployment orchestration",
+    "vercel-composition-patterns": "frontend development",
+    "vercel-deploy": "deployment orchestration",
+    "vercel-optimize": "cloud cost optimization",
+    "vercel-react-best-practices": "code refactoring and optimization",
+    "vercel-react-native-skills": "cross-platform mobile development",
+    "vercel-react-view-transitions": "web animation",
+    "verification-before-completion": "quality evaluation",
+    "verify-and-stop": "software testing",
+    video: "video",
+    "video-ad-specs": "ad creative production",
+    "video-edit": "video editing",
+    "video-extend": "video editing",
+    "video-inpainting": "video editing",
+    "video-outpainting": "video editing",
+    waapi: "web animation",
+    "wait-what": "text paraphrasing",
+    "wan-2-7": "text to video",
+    wayfinder: "task decomposition",
+    "web-artifacts-builder": "frontend development",
+    "web-design-guidelines": "design",
+    "web-perf": "code refactoring and optimization",
+    "web-search": "web search research",
+    "webapp-testing": "end-to-end and ui testing",
+    "website-to-hyperframes": "video editing",
+    "website-to-video": "video editing",
+    "wireframe-portal-title": "video editing",
+    wizard: "cli and terminal app development",
+    "workers-best-practices": "serverless function development",
+    "workers-profiling": "performance monitoring",
+    wrangler: "deployment orchestration",
+    "write-a-skill": "agent tool building",
+    "write-swift": "native mobile development",
+    "writing-beats": "content writing",
+    "writing-for-agents": "technical writing",
+    "writing-fragments": "content writing",
+    "writing-great-skills": "agent tool building",
+    "writing-guidelines": "copy editing",
+    "writing-plans": "task decomposition",
+    "writing-shape": "article and blog writing",
+    "writing-skills": "agent tool building",
+    "xcode-project-setup": "dependency management",
+    xlsx: "spreadsheet authoring",
+    "youtube-thumbnail-design": "image generation",
+    "zoom-out": "code comprehension"
+  }
+};
+
+// ../hook/src/skill-failure.ts
+var MISSING_KEY = /\b(api[ _-]?key|access[ _-]?token|credentials?)\b.{0,40}\b(missing|not set|not found|required|invalid|expired)\b|(?<!\bnot )\b(missing|no|invalid|expired)\b.{0,20}\b(api[ _-]?key|access[ _-]?token|credentials?)\b|\b(HTTP|status)(\/[\d.]+| code)?[ :=]*(401|403)\b|\b(401|403) (unauthori[sz]ed|forbidden)\b|\bunauthori[sz]ed\b|not authenticated|not logged in/i;
+var MISSING_KEY_VARIABLE = /\b[A-Z0-9_]*(API_KEY|TOKEN|SECRET)\b.{0,30}\b(not set|missing|required|unset)\b/;
+var MISSING_PROGRAM = /command not found|executable file not found|\bnot installed\b|No module named|spawn \S+ ENOENT|Cannot find module/i;
+var MISSING_FILE = /No such file or directory|\bENOENT\b|file not found|\b(file|path|directory|folder)\b(?!\.).{0,80}\bdoes not exist\b/i;
+function failureReason(text) {
+  if (MISSING_PROGRAM.test(text))
+    return "missing-program";
+  if (MISSING_KEY.test(text) || MISSING_KEY_VARIABLE.test(text))
+    return "missing-key";
+  if (MISSING_FILE.test(text))
+    return "missing-file";
+  return null;
+}
+var PROGRAM_IN_ERROR = [
+  /command not found: ([^\s'"`]+)/gi,
+  /(?:^|[\s:])([^\s:'"`]+): (?:command )?not found\b/gim,
+  /\bexec(?:utable)?:? "?([^\s"]+)"?: executable file not found/gi,
+  /\bspawn ([^\s'"`]+) ENOENT\b/gi,
+  /No module named '?([A-Za-z0-9_.-]+)'?/g,
+  /Cannot find module '([^']+)'/g,
+  /(?:^|\s)'?([^\s'"`]+)'? is not installed\b/gim
+];
+var KEY_IN_ERROR = [
+  /\b([A-Z][A-Z0-9_]{2,63})\b[^=\n]{0,30}\b(?:not set|missing|required|unset|is empty)\b/g,
+  /\b(?:set|export|provide|missing|define)\s+(?:the\s+)?(?:environment variable\s+)?\$?([A-Z][A-Z0-9_]{2,63})\b/gi,
+  /\$?\b([A-Z][A-Z0-9_]*(?:API_KEY|TOKEN|SECRET|ACCESS_KEY)[A-Z0-9_]*)\b/g
+];
+function needOf(reason, text, markdown) {
+  const kind = reason === "missing-program" ? "program" : reason === "missing-key" ? "key" : null;
+  if (kind === null)
+    return null;
+  for (const pattern of kind === "program" ? PROGRAM_IN_ERROR : KEY_IN_ERROR) {
+    for (const match of text.matchAll(pattern)) {
+      const named = match[1] ?? "";
+      const name = kind === "program" ? (named.split("/").pop() ?? "").toLowerCase() : named;
+      if (!/[a-z]/i.test(name))
+        continue;
+      const kept = publishableNeed3(name, markdown, kind);
+      if (kept !== null)
+        return kept;
+    }
+  }
+  return null;
+}
+var RUNNERS2 = new Set(["bash", "bun", "bunx", "node", "npx", "python", "python3", "sh", "uv", "uvx", "zsh"]);
+function programsOf(command) {
+  const found = [];
+  for (const part of command.split(/&&|\|\||[;|\n]/)) {
+    const words = part.trim().split(/\s+/).filter((word) => !/^[A-Za-z_][A-Za-z0-9_]*=/.test(word));
+    const first2 = words[0];
+    if (first2 === undefined || first2 === "")
+      continue;
+    const program = first2.split("/").pop() ?? first2;
+    if (!RUNNERS2.has(program)) {
+      found.push(program);
+      continue;
+    }
+    const target = words.slice(1).find((word) => !word.startsWith("-") && word !== "run");
+    if (target !== undefined)
+      found.push(target.split("/").pop() ?? target);
+  }
+  return found;
+}
+var escaped2 = (text) => text.replace(/[.*+?^${}()|[\]\\]/g, "\\$&");
+function askedBySkill(command, markdown, directory) {
+  if (new RegExp(`${escaped2(directory)}(?=[/\\s'"]|$)`).test(command))
+    return true;
+  const { cli, runs } = readRequires(markdown);
+  const named = new Set([...cli, ...runs]);
+  return programsOf(command).some((program) => named.has(program) || program.includes(".") && new RegExp(`(?:^|[\\s\`'"/(])${escaped2(program)}(?=[\\s\`'")]|$)`, "m").test(markdown));
+}
+var NODES = skill_nodes_13579_default.nodes;
+var WORDS = { labels: NODES, aliases: domain_aliases_default.aliases };
+var SKILL_NODE = skill_nodes_13579_default.skills;
+function lastUserWords(transcript) {
+  let words = "";
+  for (const line of transcript.split(`
+`)) {
+    if (!line.includes('"user"'))
+      continue;
+    try {
+      const entry = JSON.parse(line);
       if (entry.type !== "user" || entry.isMeta === true)
         continue;
       const content = entry.message?.content;
@@ -753,8 +3833,8 @@ function lastUserWords(transcript) {
   }
   return words;
 }
-function contextOf(words, skill, description) {
-  return inferDomain(words, DOMAINS) ?? classifySkill(skill.split(":").pop() ?? skill, description ?? "");
+function contextOf(words, skill) {
+  return domainInText2(words, WORDS) ?? SKILL_NODE[skill.split(":").pop() ?? skill] ?? null;
 }
 function answerOf(response) {
   const blocks = Array.isArray(response) ? response : typeof response === "object" && response !== null && Array.isArray(response.content) ? response.content : [response];
@@ -779,7 +3859,7 @@ function recordSearch(call, cwd) {
   const answer = answerOf(call.response);
   if (answer === null || !Array.isArray(answer["skills"]))
     return;
-  const domain = typeof answer["domain"] === "string" && DOMAIN.test(answer["domain"]) ? answer["domain"] : null;
+  const domain = typeof answer["domain"] === "string" && NODES.includes(answer["domain"]) ? answer["domain"] : null;
   const skills = answer["skills"].flatMap((row) => typeof row?.name === "string" && typeof row.source === "string" ? [{ name: row.name, source: row.source }] : []);
   if (skills.length === 0)
     return;
@@ -815,6 +3895,7 @@ var readIfThere = (path) => existsSync(path) ? readFileSync2(path, "utf8") : nul
 function installedSource(name, cwd) {
   const known = [
     ...installedFromSkillLock(readIfThere(join3(homedir2(), ".agents", ".skill-lock.json"))),
+    ...installedFromSkillLock(readIfThere(join3(cwd, "skills-lock.json"))),
     ...installedFromReceipts(readIfThere(join3(arpHome(), "receipts.jsonl"))),
     ...installedFromReceipts(readIfThere(join3(cwd, ".arp", "receipts.jsonl")))
   ];
@@ -865,13 +3946,13 @@ function appendFailure(line) {
 function recordSkillFailure(call, cwd) {
   const classified = classifyCall(call);
   if (classified === null || classified.outcome === "ok" || call.session === null)
-    return;
+    return null;
   const command = call.input?.["command"];
   if (typeof command !== "string")
-    return;
+    return null;
   const reason = failureReason(failureText(call));
   if (reason === null)
-    return;
+    return null;
   try {
     const log = process.env["ARP_SESSION_LOG"] ?? join3(cwd, ".arp", "session.jsonl");
     for (const skill of skillsOfSession(log, call.session)) {
@@ -883,33 +3964,258 @@ function recordSkillFailure(call, cwd) {
         continue;
       const search = foundBySearch(cwd, call.session, skill.name);
       const words = call.transcript !== null && existsSync(call.transcript) ? lastUserWords(readFileSync2(call.transcript, "utf8")) : "";
-      const domain = search?.domain ?? contextOf(words, skill.name, readRequires(markdown).description);
-      appendFailure({
+      const domain = search?.domain ?? contextOf(words, skill.name);
+      const need = needOf(reason, failureText(call), markdown);
+      const line = {
         at: new Date().toISOString(),
         session: call.session,
         skill: skill.name,
         source: skill.source ?? search?.source ?? installedSource(skill.name.split(":").pop() ?? skill.name, cwd),
         ...domain === null ? {} : { domain },
-        reason
-      });
-      return;
+        reason,
+        ...need === null ? {} : { need }
+      };
+      return appendFailure(line) ? line : null;
     }
   } catch {}
+  return null;
+}
+
+// ../hook/src/failure-answer.ts
+var ANSWER_MS = 3000;
+var WHERE = "Intuition testnet 13579";
+var IN_CONTEXT_OF = _13579_default.arp.endpointBinding.inContextOf;
+var PAYERS = [_13579_default.arp.deployer, _13579_default.arp.sponsor];
+var sha256 = (text) => createHash("sha256").update(text, "utf8").digest();
+async function answerFailure(failure, env = process.env, fetcher) {
+  if (env["ARP_FAILURE_LINES"] === "off" || failure.source === null)
+    return null;
+  const url = skillUrlOf2(failure.source, failure.skill);
+  if (url === null)
+    return null;
+  const { uri, label } = skillAtom(url, sha256);
+  const graphqlUrl = env["INTUITION_GRAPHQL_URL"] ?? _13579_default.chain.graphqlUrl;
+  let timer;
+  const late = new Promise((resolve) => {
+    timer = setTimeout(() => resolve(null), ANSWER_MS);
+  });
+  const read = readSkillLines(uri, {
+    graphqlUrl,
+    inContextOf: IN_CONTEXT_OF,
+    payers: PAYERS,
+    timeoutMs: ANSWER_MS,
+    ...fetcher === undefined ? {} : { fetch: fetcher }
+  }).then((usage) => linesSentence(usage, label, WHERE)).catch(() => null);
+  try {
+    return await Promise.race([read, late]);
+  } finally {
+    clearTimeout(timer);
+  }
+}
+function failureOutput(context) {
+  return JSON.stringify({ hookSpecificOutput: { hookEventName: "PostToolUseFailure", additionalContext: context } });
+}
+var RECORD = Symbol("record");
+async function postToolUseFailure(payload) {
+  const call = toolCallOf(payload);
+  if (call === null || call.tool !== "Bash")
+    return RECORD;
+  const failure = recordSkillFailure(call, call.cwd ?? process.cwd());
+  const context = failure === null ? null : await answerFailure(failure);
+  return context === null ? null : failureOutput(context);
+}
+
+// ../hook/src/worked.ts
+import { existsSync as existsSync2, mkdirSync as mkdirSync3, readFileSync as readFileSync3 } from "node:fs";
+import { dirname as dirname3, join as join4 } from "node:path";
+
+// ../../packages/protocol/src/local-outcomes.ts
+var OUTCOME_ANSWERS2 = ["worked", "failed", "unsure"];
+function isOutcomeAnswer2(word) {
+  return OUTCOME_ANSWERS2.includes(word);
+}
+var ANSWER = /^[\s>*`-]*arp:\s*([A-Za-z0-9._:@/-]+)\s+(worked|failed|unsure)[\s.`*]*$/gim;
+function answersIn2(text, asked) {
+  const found = new Map;
+  for (const match of text.matchAll(ANSWER)) {
+    const [, skill, word] = match;
+    if (skill === undefined || word === undefined)
+      continue;
+    const answer = word.toLowerCase();
+    if (!asked.includes(skill) || found.has(skill) || !isOutcomeAnswer2(answer))
+      continue;
+    found.set(skill, answer);
+  }
+  return found;
+}
+
+// ../hook/src/worked.ts
+var MAX_ASKED = 5;
+function isRecord5(value) {
+  return typeof value === "object" && value !== null && !Array.isArray(value);
+}
+var field = (raw, snake, camel) => raw[snake] ?? raw[camel];
+var text = (value) => typeof value === "string" && value.length > 0 ? value : null;
+function stopInputOf(raw) {
+  if (!isRecord5(raw) || field(raw, "hook_event_name", "hookEventName") !== "Stop")
+    return null;
+  const session = text(field(raw, "session_id", "sessionId"));
+  if (session === null)
+    return null;
+  return {
+    session,
+    cwd: text(raw["cwd"]) ?? process.cwd(),
+    active: field(raw, "stop_hook_active", "stopHookActive") === true,
+    last: text(field(raw, "last_assistant_message", "lastAssistantMessage")) ?? "",
+    transcript: text(field(raw, "transcript_path", "transcriptPath"))
+  };
+}
+function questionFor(skills) {
+  return "ARP asks once per skill: did each skill you used here do its job, and did the task finish with it? " + "Reply with one line per skill and nothing else: `arp: <skill> worked`, `arp: <skill> failed` or `arp: <skill> unsure`. " + `Kept on this machine unless \`arp share\` is on. Skills: ${skills.join(", ")}`;
+}
+var isArpsOwn = (name) => name === "arp" || name.endsWith(":arp");
+function rows(path) {
+  if (!existsSync2(path))
+    return [];
+  const found = [];
+  for (const line of readFileSync3(path, "utf8").split(`
+`)) {
+    if (line.trim() === "")
+      continue;
+    try {
+      const row = JSON.parse(line);
+      if (isRecord5(row))
+        found.push(row);
+    } catch {}
+  }
+  return found;
+}
+function withOf(skill, source, cwd, before) {
+  const name = skill.split(":").pop() ?? skill;
+  let need = null;
+  let latest = "";
+  for (const row of rows(join4(arpHome(), "failures.jsonl"))) {
+    const { skill: rowSkill, source: rowSource, need: rowNeed, at } = row;
+    if (typeof rowSkill !== "string" || typeof rowNeed !== "string" || typeof at !== "string")
+      continue;
+    if (rowSkill !== skill && rowSkill !== name && !rowSkill.endsWith(`:${name}`))
+      continue;
+    if (typeof rowSource === "string" && source !== null && rowSource !== source)
+      continue;
+    if (at > before || at < latest)
+      continue;
+    need = rowNeed;
+    latest = at;
+  }
+  if (need === null)
+    return null;
+  const directory = skillDirectory(name, cwd);
+  if (directory === null)
+    return null;
+  return publishableNeed3(need, readFileSync3(join4(directory, "SKILL.md"), "utf8"));
+}
+function record(input, asked, outcome, at) {
+  const search = foundBySearch(input.cwd, input.session, asked.skill);
+  const words = input.transcript !== null && existsSync2(input.transcript) ? lastUserWords(readFileSync3(input.transcript, "utf8")) : "";
+  const domain = search?.domain ?? contextOf(words, asked.skill);
+  const source = asked.source ?? search?.source ?? installedSource(asked.skill.split(":").pop() ?? asked.skill, input.cwd);
+  const named = outcome === "worked" ? withOf(asked.skill, source, input.cwd, at) : null;
+  const line = {
+    at,
+    session: input.session,
+    skill: asked.skill,
+    source,
+    ...domain === null ? {} : { domain },
+    outcome,
+    ...named === null ? {} : { with: named }
+  };
+  appendOwn(join4(arpHome(), "outcomes.jsonl"), `${JSON.stringify(line)}
+`);
+}
+async function stop(payload) {
+  const input = stopInputOf(payload);
+  if (input === null)
+    return null;
+  const ask = onStop(input);
+  return ask === null ? null : JSON.stringify(ask);
+}
+function onStop(input, now = new Date) {
+  try {
+    const home = arpHome();
+    mkdirSync3(home, { recursive: true });
+    const askedPath = join4(home, "asked.jsonl");
+    const asked = rows(askedPath).flatMap((row) => row["session"] === input.session && typeof row["skill"] === "string" ? [{
+      at: typeof row["at"] === "string" ? row["at"] : "",
+      session: input.session,
+      skill: row["skill"],
+      source: typeof row["source"] === "string" ? row["source"] : null
+    }] : []);
+    const answered = new Set(rows(join4(home, "outcomes.jsonl")).flatMap((row) => row["session"] === input.session && typeof row["skill"] === "string" ? [row["skill"]] : []));
+    const pending = asked.filter((row) => !answered.has(row.skill));
+    const at = now.toISOString();
+    for (const [skill, answer] of answersIn2(input.last, pending.map((row) => row.skill))) {
+      const row = pending.find((entry) => entry.skill === skill);
+      if (row !== undefined)
+        record(input, row, answer, at);
+    }
+    if (input.active)
+      return null;
+    const known = new Set(asked.map((row) => row.skill));
+    const log = process.env["ARP_SESSION_LOG"] ?? join4(input.cwd, ".arp", "session.jsonl");
+    const fresh = [];
+    for (const skill of skillsOfSession(log, input.session)) {
+      if (known.has(skill.name) || isArpsOwn(skill.name))
+        continue;
+      known.add(skill.name);
+      fresh.push({ at, session: input.session, skill: skill.name, source: skill.source });
+    }
+    const ask = fresh.slice(0, MAX_ASKED);
+    if (ask.length === 0)
+      return null;
+    mkdirSync3(dirname3(askedPath), { recursive: true });
+    appendOwn(askedPath, ask.map((row) => `${JSON.stringify(row)}
+`).join(""));
+    return { decision: "block", reason: questionFor(ask.map((row) => row.skill)) };
+  } catch {
+    return null;
+  }
 }
 
 // ../hook/src/hook.ts
+var ROUTES = {
+  PreToolUse: (payload) => preToolUse(payload),
+  PostToolUseFailure: (payload) => postToolUseFailure(payload),
+  Stop: (payload) => stop(payload)
+};
+async function route(handler, payload) {
+  let out = null;
+  try {
+    out = await handler(payload);
+  } catch {
+    out = null;
+  }
+  if (out === RECORD)
+    return;
+  if (out === null)
+    process.exit(0);
+  process.stdout.write(`${out}
+`, () => process.exit(0));
+  await new Promise(() => {
+    return;
+  });
+}
 function fetchedRefs(cwd) {
-  const path = process.env["ARP_RECEIPTS"] ?? join4(cwd, ".arp", "receipts.jsonl");
-  if (!existsSync2(path))
+  const path = process.env["ARP_RECEIPTS"] ?? join5(cwd, ".arp", "receipts.jsonl");
+  if (!existsSync3(path))
     return [];
-  let text;
+  let text2;
   try {
-    text = readFileSync3(path, "utf8");
+    text2 = readFileSync4(path, "utf8");
   } catch {
     return [];
   }
-  const rows = [];
-  for (const line of text.split(`
+  const rows2 = [];
+  for (const line of text2.split(`
 `)) {
     if (line.trim().length === 0)
       continue;
@@ -917,7 +4223,7 @@ function fetchedRefs(cwd) {
       const row = JSON.parse(line);
       if (typeof row.skill !== "string" || typeof row.atom !== "string")
         continue;
-      rows.push({
+      rows2.push({
         skill: row.skill,
         atom: row.atom,
         url: typeof row.url === "string" ? row.url : "",
@@ -927,7 +4233,7 @@ function fetchedRefs(cwd) {
       continue;
     }
   }
-  return rows;
+  return rows2;
 }
 function responseOf(call, outcome) {
   if (!call.hasResponse)
@@ -941,7 +4247,7 @@ function responseOf(call, outcome) {
   }
   return { answered: raw !== null, bytes, failed: outcome !== "ok" };
 }
-function readString(from, key) {
+function readString2(from, key) {
   const value = from?.[key];
   return typeof value === "string" && value.trim().length > 0 ? value.trim() : null;
 }
@@ -957,15 +4263,15 @@ function serverRef(server, cwd) {
 }
 function whatOf(call, cwd, fetched) {
   if (call.tool === "Skill") {
-    const name = readString(call.input, "skill");
+    const name = readString2(call.input, "skill");
     if (name === null)
       return null;
-    const source = readString(call.input, "source");
+    const source = readString2(call.input, "source");
     return {
       kind: "skill",
       name,
       source,
-      ref: readString(call.input, "atom_id") ?? readString(call.input, "tool_atom_id") ?? readString(call.input, "url") ?? refForInvocation(name, source, fetched())
+      ref: readString2(call.input, "atom_id") ?? readString2(call.input, "tool_atom_id") ?? readString2(call.input, "url") ?? refForInvocation(name, source, fetched())
     };
   }
   const mcp = /^mcp__([^_]+(?:_[^_]+)*?)__/.exec(call.tool);
@@ -984,7 +4290,11 @@ try {
   process.exit(0);
 }
 if (process.argv.includes("--capture"))
-  capture(parsed, join4(homedir3(), ".arp", "captures"));
+  capture(parsed, join5(homedir3(), ".arp", "captures"));
+var event = eventOf(parsed);
+var handler = event === null ? undefined : ROUTES[event];
+if (handler !== undefined)
+  await route(handler, parsed);
 var call = toolCallOf(parsed);
 if (call === null)
   process.exit(0);
@@ -1001,8 +4311,8 @@ if (what === null)
 var classified = classifyCall(call);
 if (classified === null)
   process.exit(0);
-var own = join4(cwd, ".arp");
-var path = process.env["ARP_SESSION_LOG"] ?? join4(own, "session.jsonl");
+var own = join5(cwd, ".arp");
+var path = process.env["ARP_SESSION_LOG"] ?? join5(own, "session.jsonl");
 var line = {
   at: new Date().toISOString(),
   session: call.session,
@@ -1015,11 +4325,11 @@ var line = {
   response: responseOf(call, classified.outcome)
 };
 try {
-  mkdirSync3(dirname3(path), { recursive: true });
-  if (dirname3(path) === own && lstatSync2(own).isSymbolicLink())
+  mkdirSync4(dirname4(path), { recursive: true });
+  if (dirname4(path) === own && lstatSync2(own).isSymbolicLink())
     throw new Error(`${own} is a symbolic link`);
-  const ignore = join4(own, ".gitignore");
-  if (dirname3(path) === own && !existsSync2(ignore))
+  const ignore = join5(own, ".gitignore");
+  if (dirname4(path) === own && !existsSync3(ignore))
     writeFileSync2(ignore, `*
 `, { flag: "wx" });
   appendOwn(path, `${JSON.stringify(line)}