@iamu-dev/iamu 0.2.0-beta.5 → 0.2.0-beta.7

package/dist/bin.js

@@ -200,17 +200,24 @@ function defaultConfig(params) {
     cloud_api_url: null
   };
 }
-var baseProjectFields, localOnlyConfig, cloudLinkedConfig, projectConfigSchema;
+var modelTierSchema, projectModelsConfigSchema, baseProjectFields, localOnlyConfig, cloudLinkedConfig, projectConfigSchema;
 var init_config = __esm({
   "../../packages/protocol/dist/schemas/frontmatter/config.js"() {
     "use strict";
     init_common();
     init_version();
+    modelTierSchema = z4.string().min(1).max(128).describe("Model role tier ('fast', 'thorough', 'inherit') or a specific model identifier.");
+    projectModelsConfigSchema = z4.object({
+      walker: modelTierSchema.optional(),
+      matcher: modelTierSchema.optional(),
+      judge: modelTierSchema.optional()
+    }).strict();
     baseProjectFields = {
       schema_version: z4.literal(SCHEMA_VERSION),
       id: uuidSchema,
       project_slug: slugSchema,
-      project_name: z4.string().min(1)
+      project_name: z4.string().min(1),
+      models: projectModelsConfigSchema.optional()
     };
     localOnlyConfig = z4.object({
       ...baseProjectFields,
@@ -817,7 +824,14 @@ var init_scenario = __esm({
       reviews: z15.array(z15.string().min(1).max(64)).max(16).default([]),
       baseline_score: z15.number().min(0).max(100).nullish(),
       current_score: z15.number().min(0).max(100).nullish(),
-      target_score: z15.number().min(0).max(100).default(95)
+      target_score: z15.number().min(0).max(100).default(95),
+      /** Environment state preconditions and approximation guidance. */
+      state: z15.object({
+        /** Description of data/records expected to exist in the environment. */
+        description: z15.string().max(2048).nullish(),
+        /** Guidance on how to squint or approximate if exact records are absent. */
+        approximation: z15.string().max(2048).nullish()
+      }).nullish()
     });
   }
 });
@@ -1154,7 +1168,21 @@ var init_walkthrough_bundle = __esm({
           judge_prompt: z19.string().optional(),
           judge_model: z19.string().optional()
         }).optional()
-      }).optional()
+      }).optional(),
+      models: z19.object({
+        walker: z19.string().optional(),
+        matcher: z19.string().optional(),
+        judge: z19.string().optional(),
+        walker_reported: z19.string().optional(),
+        matcher_reported: z19.string().optional(),
+        judge_reported: z19.string().optional(),
+        walker_tier: z19.string().optional(),
+        matcher_tier: z19.string().optional(),
+        judge_tier: z19.string().optional(),
+        execution_mode: z19.enum(["multi_agent", "single_agent"]).optional(),
+        velocity_sec_per_step: z19.number().optional(),
+        total_duration_sec: z19.number().optional()
+      }).optional().describe("AI model configuration and reported runtime provenance for this walkthrough.")
     });
     walkthroughBundleSchema = z19.object({
       manifest: bundleManifestSchema,
@@ -1377,6 +1405,7 @@ var init_walkthrough_canonical = __esm({
       build: buildMetadataSchema.optional(),
       viewport: runViewportSchema.optional(),
       environment: z20.string().optional(),
+      models: z20.record(z20.unknown()).optional(),
       findings: z20.array(canonicalFindingSchema),
       steps: z20.array(canonicalStepSchema),
       human_tests: z20.array(canonicalHumanTestSchema),
@@ -7248,6 +7277,7 @@ async function buildCanonical(projectRoot, manifest, completedAt, inputs) {
     ...manifest.build ? { build: manifest.build } : {},
     ...manifest.viewport ? { viewport: manifest.viewport } : {},
     ...manifest.environment ? { environment: manifest.environment } : {},
+    ...manifest.models ? { models: manifest.models } : {},
     findings: canonicalFindings,
     steps: canonicalSteps,
     human_tests: humanTests,
@@ -7408,6 +7438,20 @@ function renderCanonicalMarkdown(canonical) {
     lines.push(`- Viewport: ${canonical.viewport.device} (${canonical.viewport.width}\xD7${canonical.viewport.height})`);
   }
   lines.push(`- Started: ${canonical.provenance.started_at}`, `- Completed: ${canonical.provenance.completed_at}`, `- Walker prompt: ${canonical.provenance.walker_prompt.id}@${canonical.provenance.walker_prompt.version}`, `- Findings: ${canonical.findings.length}`, `- Steps: ${canonical.steps.length}`);
+  const models = canonical.models;
+  if (models) {
+    const parts = [
+      models.walker ? `Walker: ${models.walker}${models.walker_reported ? ` (${models.walker_reported})` : ""}` : "",
+      models.judge ? `Judge: ${models.judge}${models.judge_reported ? ` (${models.judge_reported})` : ""}` : "",
+      models.matcher ? `Matcher: ${models.matcher}` : ""
+    ].filter(Boolean);
+    if (parts.length > 0) {
+      lines.push(`- Models: ${parts.join(" \xB7 ")}`);
+    }
+    if (typeof models.velocity_sec_per_step === "number") {
+      lines.push(`- Velocity: ~${models.velocity_sec_per_step}s/step (${models.total_duration_sec}s total)`);
+    }
+  }
   const what = canonical.what_worked;
   if (what && what.length > 0) {
     lines.push("", "## What worked");
@@ -7777,6 +7821,7 @@ async function assembleWalkthroughBundle(projectRoot, runId) {
       report_markdown: report3,
       verdict: verdictArray,
       judge_meta: judgeMeta,
+      ...manifest.models ? { models: manifest.models } : {},
       ...buildFields(manifest),
       ...await readCanonicalExtras(projectRoot, runId)
     },
@@ -12287,15 +12332,6 @@ function mcpJson(cmd) {
     2
   );
 }
-function vscodeMcpJson(cmd) {
-  return JSON.stringify(
-    {
-      servers: { iamu: { type: "stdio", command: cmd[0], args: cmd.slice(1) } }
-    },
-    null,
-    2
-  );
-}
 function detectHosts(projectRoot) {
   return HOST_ADAPTERS.filter(
     (host) => host.detectionPaths.some((p) => existsSync3(join3(projectRoot, p)))
@@ -12344,13 +12380,17 @@ var init_registry = __esm({
       {
         name: "GitHub Copilot",
         value: "github-copilot",
-        detectionPaths: [".vscode", ".github/copilot-instructions.md"],
-        supportsSkills: false,
-        getSkillsDir: () => "",
-        getCommandPath: (root, id) => join3(root, ".github", "prompts", `iamu-${id}.prompt.md`),
-        formatCommand: descOnlyFormat,
-        getMcpConfigPath: (root) => join3(root, ".vscode", "mcp.json"),
-        formatMcpConfig: vscodeMcpJson
+        detectionPaths: [
+          ".github/copilot-instructions.md",
+          ".github/skills",
+          ".copilot"
+        ],
+        supportsSkills: true,
+        getSkillsDir: (root) => join3(root, ".github", "skills"),
+        getCommandPath: (root, id) => join3(root, ".github", "skills", `iamu-${id}`, "SKILL.md"),
+        formatCommand: richFormat,
+        getMcpConfigPath: (root) => join3(root, ".mcp.json"),
+        formatMcpConfig: mcpJson
       },
       {
         name: "Cline",
@@ -12532,6 +12572,14 @@ function isCliWrittenSkill(path28) {
     return false;
   }
 }
+function isCliWrittenPrompt(path28) {
+  try {
+    const content = readFileSync3(path28, "utf-8");
+    return content.includes("iamu") || /^generatedBy:/m.test(content);
+  } catch {
+    return false;
+  }
+}
 function planSkillFiles(hosts, projectRoot, version2) {
   const skillTemplates = loadSkillTemplates(version2);
   const files = [];
@@ -12560,6 +12608,19 @@ function planSkillFiles(hosts, projectRoot, version2) {
           }
         }
       }
+      if (host.value === "github-copilot") {
+        for (const cmd of commandsFromSkills(skillTemplates)) {
+          const legacyPrompt = join4(
+            projectRoot,
+            ".github",
+            "prompts",
+            `iamu-${cmd.id}.prompt.md`
+          );
+          if (isCliWrittenPrompt(legacyPrompt)) {
+            files.push({ host, path: legacyPrompt, content: null });
+          }
+        }
+      }
     } else {
       for (const cmd of commandsFromSkills(skillTemplates)) {
         files.push({
@@ -12584,6 +12645,15 @@ function detectInstalledHosts(projectRoot) {
       const skillDir = join4(host.getSkillsDir(projectRoot), "iamu-setup");
       if (existsSync5(skillDir)) return true;
     }
+    if (host.value === "github-copilot") {
+      const legacyPrompt = join4(
+        projectRoot,
+        ".github",
+        "prompts",
+        "iamu-setup.prompt.md"
+      );
+      if (existsSync5(legacyPrompt)) return true;
+    }
     return false;
   });
 }
@@ -14434,6 +14504,18 @@ async function startRun(scenarioSlug, opts) {
   }
   const build = detectBuild({ cwd: projectRoot, label: opts.build });
   const environment = resolveEnvironment(opts.env, start2.url);
+  const cfg = await readConfig(projectRoot).catch(() => null);
+  const walker = opts.walker ?? process.env.IAMU_WALKER_MODEL ?? cfg?.models?.walker ?? "fast";
+  const judge = opts.judge ?? process.env.IAMU_JUDGE_MODEL ?? cfg?.models?.judge ?? "thorough";
+  const matcher = opts.matcher ?? process.env.IAMU_MATCHER_MODEL ?? cfg?.models?.matcher ?? "thorough";
+  const models = {
+    walker,
+    judge,
+    matcher,
+    walker_tier: ["fast", "thorough", "inherit"].includes(walker) ? walker : "custom",
+    judge_tier: ["fast", "thorough", "inherit"].includes(judge) ? judge : "custom",
+    matcher_tier: ["fast", "thorough", "inherit"].includes(matcher) ? matcher : "custom"
+  };
   let revisit = [];
   try {
     revisit = await computeRevisitPlaces(projectRoot, scenarioSlug, start2.url);
@@ -14451,6 +14533,7 @@ async function startRun(scenarioSlug, opts) {
     status: "running",
     created_at: (/* @__PURE__ */ new Date()).toISOString(),
     completed_at: null,
+    models,
     ...build ? { build } : {},
     ...environment ? { environment } : {},
     ...revisit.length > 0 ? { revisit } : {},
@@ -14472,6 +14555,7 @@ async function startRun(scenarioSlug, opts) {
           ...targetWarning ? { target_warning: targetWarning } : {},
           persona_slug: fm.persona,
           outline_slug: fm.outline,
+          models,
           build: build ?? null,
           environment: environment ?? null,
           revisit,
@@ -14490,6 +14574,9 @@ async function startRun(scenarioSlug, opts) {
   console.log(`  ${chalk11.dim("persona:")}  ${fm.persona}`);
   console.log(`  ${chalk11.dim("outline:")}  ${fm.outline}`);
   console.log(`  ${chalk11.dim("target:")}   ${start2.url}`);
+  console.log(
+    `  ${chalk11.dim("models:")}   walker: ${chalk11.cyan(walker)} \xB7 judge: ${chalk11.cyan(judge)} \xB7 matcher: ${chalk11.cyan(matcher)}`
+  );
   if (targetWarning) {
     console.log(`  ${symbols.warning} ${chalk11.yellow(targetWarning)}`);
   }
@@ -16286,7 +16373,10 @@ var init_browse_actions = __esm({
         usage: "goto <url>",
         minArgs: 1,
         run: async (p, args) => {
-          await p.goto(args[0], { waitUntil: "domcontentloaded" });
+          await p.goto(args[0], {
+            waitUntil: "domcontentloaded",
+            timeout: 1e4
+          });
         }
       },
       click: {
@@ -16447,6 +16537,12 @@ var init_browse_driver = __esm({
       commandLine: (pid) => readProcessCommandLine(pid),
       kill(pid) {
         try {
+          if (process.platform !== "win32") {
+            try {
+              process.kill(-pid, "SIGTERM");
+            } catch {
+            }
+          }
           process.kill(pid, "SIGTERM");
         } catch {
         }
@@ -16673,6 +16769,7 @@ __export(browse_exports, {
   readBrowseState: () => readBrowseState,
   shouldDisableSandbox: () => shouldDisableSandbox,
   start: () => start,
+  waitForVisualStability: () => waitForVisualStability,
   withPage: () => withPage
 });
 import { randomUUID as randomUUID7 } from "crypto";
@@ -16822,6 +16919,8 @@ function chromeArgs(opts, port, headless) {
     `--window-size=${opts.width ?? "1280"},${opts.height ?? "900"}`,
     "--no-first-run",
     "--no-default-browser-check",
+    "--password-store=basic",
+    "--use-mock-keychain",
     ...headless ? ["--headless=new", "--disable-dev-shm-usage", "--disable-gpu"] : [],
     ...shouldDisableSandbox() ? ["--no-sandbox"] : [],
     "about:blank"
@@ -16879,6 +16978,45 @@ async function stop(opts, driver) {
   }
   report2(opts, { stopped: true }, [`  ${chalk20.green("\u2713")} Browser stopped`]);
 }
+async function waitForVisualStability(page, maxWaitMs = 300) {
+  let timer;
+  try {
+    const nodeTimeout = new Promise((resolve4) => {
+      timer = setTimeout(resolve4, maxWaitMs + 100);
+    });
+    const browserWait = page.evaluate((maxWait) => {
+      return new Promise((resolve4) => {
+        const timeout = setTimeout(resolve4, maxWait);
+        const done = () => {
+          clearTimeout(timeout);
+          resolve4();
+        };
+        const g = globalThis;
+        if (typeof g.requestAnimationFrame !== "function") {
+          done();
+          return;
+        }
+        g.requestAnimationFrame(() => {
+          g.requestAnimationFrame?.(() => {
+            const animations = g.document?.getAnimations ? g.document.getAnimations().filter((a) => a.playState === "running" && a.finished) : [];
+            if (animations.length === 0) {
+              done();
+              return;
+            }
+            Promise.allSettled(
+              animations.map((a) => a.finished)
+            ).finally(done);
+          });
+        });
+      });
+    }, maxWaitMs).catch(() => {
+    });
+    await Promise.race([browserWait, nodeTimeout]);
+  } catch {
+  } finally {
+    if (timer) clearTimeout(timer);
+  }
+}
 async function withPage(opts, driver, fn, target) {
   const state = readBrowseState();
   if (!state) {
@@ -16898,9 +17036,9 @@ async function withPage(opts, driver, fn, target) {
     await fn(page);
     await page.waitForLoadState("domcontentloaded").catch(() => {
     });
-    await page.waitForTimeout(400);
     let screenshot = null;
     if (opts.shot) {
+      await waitForVisualStability(page);
       screenshot = path22.resolve(opts.shot);
       mkdirSync(path22.dirname(screenshot), { recursive: true });
       await page.screenshot({ path: screenshot, fullPage: Boolean(opts.full) });
@@ -18459,13 +18597,21 @@ function defaultFrontmatter(type2, id, slug, name, opts) {
         url: opts.url,
         authority_level: opts.authority ?? "informational"
       };
-    case "scenario":
+    case "scenario": {
+      const stateObj = opts.stateDescription || opts.stateApproximation ? {
+        state: {
+          ...opts.stateDescription ? { description: opts.stateDescription } : {},
+          ...opts.stateApproximation ? { approximation: opts.stateApproximation } : {}
+        }
+      } : {};
       return {
         ...base,
         outline: opts.outline ?? null,
         persona: opts.persona ?? null,
-        ...opts.entry ? { entry_path: opts.entry } : {}
+        ...opts.entry ? { entry_path: opts.entry } : {},
+        ...stateObj
       };
+    }
     default:
       return base;
   }
@@ -18859,6 +19005,17 @@ function composeWalkerPrompt(p) {
     "",
     "## First-person brief",
     p.brief ?? "",
+    ...p.context ? ["", "## Product domain & regulatory context", p.context.trim()] : [],
+    ...p.state?.description || p.state?.approximation ? [
+      "",
+      "## Expected state & semantic squinting",
+      p.state.description ? `Expected state: ${p.state.description}` : "",
+      p.state.approximation ? `Approximation strategy: ${p.state.approximation}` : "",
+      "Squinting Guardrails:",
+      "\u2022 If the exact target record/option is not available, apply the approximation strategy above to pick the closest plausible entity within the domain flow.",
+      "\u2022 Squinting does NOT permit drifting to unrelated settings, account profiles, or off-target pages. Stay strictly within the scenario journey.",
+      "\u2022 If the system is completely empty or unseeded, and your persona role cannot create data, do NOT report false product bugs about the empty state. Note that preconditions are unseeded/blocked."
+    ].filter(Boolean) : [],
     ...revisitSection(p.revisit ?? []),
     "",
     "## Access",
@@ -18887,7 +19044,7 @@ function composeWalkerPrompt(p) {
     `Screenshots to: ${rd}/screenshots/step-N.png \u2014 absolute paths.`,
     `Record findings with: ${iamu} new-finding --scenario ${p.scenarioSlug} --run ${p.runId} --step <N> --title "..." --type <helps|hinders|missing|irrelevant|incorrect> --severity <critical|major|minor|observation> --ui-fact "..." --consequence "..." --task-impact <blocks|impairs|slows|neutral|accelerates|enables> --confidence <high|medium|low> --json \u2014 then add the body (description + "## Recommendation") below the frontmatter of the file it returns.`,
     "`--step <N>` is the step whose screenshot shows the problem (or strength): it is the evidence people see next to the finding, so record that step and its screenshot first, and pick the screenshot where the thing is actually visible.",
-    `Before you finish, write ${rd}/summary.json: {"what_worked": [...], "open_questions": [...]} \u2014 one sentence each, in the persona's language.`,
+    `Before you finish, write ${rd}/summary.json: {"what_worked": [...], "open_questions": [...], "model": "<your model identifier if known, e.g. claude-3-5-haiku, gemini-3.8-flash, or unspecified>"} \u2014 one sentence each, in the persona's language.`,
     "",
     "Do not read the product's source code, other iamu/ entity files, or earlier runs: you know only what the persona knows.",
     "",
@@ -19007,7 +19164,8 @@ async function instructionsWalk(opts) {
   const result = {
     scenario: {
       slug: sfm.slug,
-      brief: scenario.body.trim() || null
+      brief: scenario.body.trim() || null,
+      state: sfm.state ?? null
     },
     persona: {
       slug: sfm.persona,
@@ -19061,7 +19219,9 @@ async function instructionsWalk(opts) {
       runDirPath,
       projectRoot,
       surface: opts.surface ?? "browse",
-      viewport
+      viewport,
+      state: sfm.state ?? null,
+      context: contextYml
     }) : null
   };
   if (opts.json) {
@@ -19073,6 +19233,11 @@ async function instructionsWalk(opts) {
   console.log(`  ${chalk26.dim("scenario:")}  ${sfm.slug}`);
   console.log(`  ${chalk26.dim("persona:")}   ${pfm.name} (${pfm.role})`);
   console.log(`  ${chalk26.dim("outline:")}   ${ofm.title}`);
+  if (sfm.state?.description || sfm.state?.approximation) {
+    console.log(
+      `  ${chalk26.dim("state:")}     ${sfm.state.description ? sfm.state.description : sfm.state.approximation ?? ""}`
+    );
+  }
   if (targetUrl) console.log(`  ${chalk26.dim("target:")}    ${targetUrl}`);
   console.log(
     `  ${chalk26.dim("access:")}    ${projectAccess || personaAccess ? `${projectAccess ? "project" : ""}${projectAccess && personaAccess ? " + " : ""}${personaAccess ? "persona" : ""}` : "not recorded (walker would run as whoever is signed in)"}`
@@ -19709,6 +19874,36 @@ async function complete(runId, opts) {
     return;
   }
   const completedAt = pickCompletedAt(run.manifest);
+  const walkerSummary = await readWalkerSummary(run.runDirPath);
+  let judgeReported;
+  try {
+    const verdictsRaw = await fs26.readFile(
+      path26.join(run.runDirPath, "verdicts.json"),
+      "utf8"
+    );
+    const parsed = JSON.parse(verdictsRaw);
+    if (parsed.provenance?.model) {
+      judgeReported = parsed.provenance.model;
+    }
+  } catch {
+  }
+  const stepCount = provenanceAudit.step_count;
+  const durationSec = provenanceAudit.duration_seconds ?? (run.manifest.created_at && Date.parse(completedAt) > Date.parse(run.manifest.created_at) ? Math.round(
+    (Date.parse(completedAt) - Date.parse(run.manifest.created_at)) / 1e3 * 10
+  ) / 10 : null);
+  const velocitySec = provenanceAudit.avg_step_interval_seconds ?? (durationSec != null && stepCount > 1 ? Math.round(durationSec / (stepCount - 1) * 10) / 10 : null);
+  const existingModels = run.manifest.models ?? {};
+  const updatedModels = {
+    ...existingModels,
+    ...walkerSummary.model ? { walker_reported: walkerSummary.model } : {},
+    ...judgeReported ? { judge_reported: judgeReported } : {},
+    ...durationSec != null ? { total_duration_sec: durationSec } : {},
+    ...velocitySec != null ? { velocity_sec_per_step: velocitySec } : {}
+  };
+  run.manifest = {
+    ...run.manifest,
+    models: updatedModels
+  };
   const assembled = await assembleRun(
     run.projectRoot,
     runId,
@@ -19869,6 +20064,17 @@ function printCompletionSummary(run, completedAt, assembled, extras, provenanceA
   console.log(`  steps      ${count(canonical.steps.length)}`);
   console.log(`  started    ${chalk30.dim(manifest.created_at)}`);
   console.log(`  completed  ${chalk30.dim(completedAt)}`);
+  const models = manifest.models;
+  if (models?.velocity_sec_per_step != null) {
+    console.log(
+      `  velocity   ${chalk30.cyan(`~${models.velocity_sec_per_step}s/step`)} ${chalk30.dim(`(${models.total_duration_sec}s total)`)}`
+    );
+  }
+  if (models && (models.walker || models.judge || models.matcher)) {
+    const w = models.walker_reported ? `${models.walker} (${models.walker_reported})` : models.walker ?? "fast";
+    const j = models.judge_reported ? `${models.judge} (${models.judge_reported})` : models.judge ?? "thorough";
+    console.log(`  models     ${chalk30.dim(`walker: ${w} \xB7 judge: ${j}`)}`);
+  }
   console.log();
   console.log(
     `  ${chalk30.dim("canonical:")} ${chalk30.dim(path26.relative(projectRoot, assembled.canonicalPath))}`
@@ -19962,9 +20168,11 @@ async function readWalkerSummary(runDirPath) {
       await fs26.readFile(path26.join(runDirPath, "summary.json"), "utf8")
     );
     const strings = (v) => Array.isArray(v) ? v.filter((x) => typeof x === "string" && x.trim() !== "") : void 0;
+    const model = typeof data.model === "string" && data.model.trim() !== "" ? data.model.trim() : void 0;
     return {
       what_worked: strings(data.what_worked),
-      open_questions: strings(data.open_questions)
+      open_questions: strings(data.open_questions),
+      model
     };
   } catch {
     return {};
@@ -20713,6 +20921,12 @@ function registerProjectCommands(program2) {
   ).option("--outline <slug>", "Outline slug (scenario only)").option("--persona <slug>", "Persona slug (scenario only)").option(
     "--entry <path>",
     "Where the situation starts, as a path on the product, e.g. /en/episodes/1 (scenario only)"
+  ).option(
+    "--state-description <text>",
+    "Assumed environment data/records state (scenario only)"
+  ).option(
+    "--state-approximation <text>",
+    "Guidance on how persona should approximate/squint if records are missing (scenario only)"
   ).option("--url <url>", "Document the source cites (source only, required)").option(
     "--authority <level>",
     "legislation | regulation | standard | company_policy | guideline | informational (source only, default informational)"
@@ -20754,6 +20968,15 @@ function registerWalkthroughCommands(program2) {
   ).option(
     "--env <name>",
     "Environment the product runs in (default: the target URL's origin)"
+  ).option(
+    "--walker <tierOrModel>",
+    "Model tier ('fast', 'thorough', 'inherit') or model ID for driving browser navigation (default: fast)"
+  ).option(
+    "--judge <tierOrModel>",
+    "Model tier ('fast', 'thorough', 'inherit') or model ID for human test evaluation (default: thorough)"
+  ).option(
+    "--matcher <tierOrModel>",
+    "Model tier ('fast', 'thorough', 'inherit') or model ID for finding matching (default: thorough)"
   ).option(
     "--check",
     "Enforce target URL reachability (fail fast if server is offline)"