worktrust 0.8.8 → 0.9.1

package/README.md

@@ -173,6 +173,9 @@ nothing it does not:
 | `unconfirmed` | checks the stretch ran whose outcome no reader could see (Codex's code mode prints no exit code; a background terminal), per kind: counted, never as a pass (0.8.5) |
 | `autonomy`, `authorship` | the permission mode the AI app ran in, on one scale (ask, edits, plan, auto, full), how often it changed and the turns in plan mode; of the stretch's commits, how many name an AI as co-author (the trailer is read here, its names never leave) (0.8.7) |
 | `framing`, `quality`, `risk`, `tool_mix`, `reads`, `context_files` | how the work was framed (your turns before the agent's first action, a plan first or not), checked (a check passing after the last change, the failures before a check passed, the change looked at before delivery; security scans count as checks), risked (destructive commands proposed, refused, run), which kinds of tools were used, files read again unchanged, and which of the AI app's own context files (instructions, skills, agents, commands, settings, MCP) the commits changed; kinds and counts only (0.8.8) |
+| `quality.change_blocks`, `checked_blocks`, `rework_cycles`, `stage` | how many change blocks (edits with no check between them) a check followed, how often a failed check was followed by another change before it ran again (rework), and the furthest step the stretch reached here: attempted, verified, committed, pushed, pr, deployed; counts and one word only (0.8.9) |
+| `timing` | how fast the stretch moved: seconds from your first turn to the first action and from the first change to the first check, the actions and seconds from a failure to the change that answered it, the calls from a failed check to it passing, and the calls back to work after a compaction; medians from the order and the clocks, never what was said (0.9.0) |
+| `hygiene` | risk and hygiene: rollbacks, privileged commands (sudo, a forced push), hooks skipped (--no-verify), secrets touched (a .env printed, a private key read), destructive commands and how many a look or a check came before, the answer to a guard (retried, changed, stopped), and whether a check followed delivery; counts only, never the command or the path (0.9.1) |
 | `collector_version` | the CLI version that measured it |
 | `seq`, `prev`, `hash` | the line's place, the hash of the line before it, and its own hash |
 

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "worktrust",
-  "version": "0.8.8",
+  "version": "0.9.1",
   "description": "Couple this computer to WorkTrust: approve a code in your browser, and every AI app here reports what you built. Metadata only, no dependencies.",
   "type": "module",
   "bin": {

package/preserve-lines.mjs

@@ -105,7 +105,23 @@ export function archiveLine(entry, behaviour = null, { ids = {}, collector, salt
   // PRACTICE (0.8.8): framing, verification quality, risk, tool mix, re-reads, context files; whole counts and flags only.
   const bool = (value) => value === true || value === false;
   if (entry.framing && whole(entry.framing.turns_before_action) && bool(entry.framing.planned_first)) line.framing = { turns_before_action: entry.framing.turns_before_action, planned_first: entry.framing.planned_first };
-  if (entry.quality && whole(entry.quality.fix_cycles_max) && (bool(entry.quality.checked_after_change) || entry.quality.checked_after_change === null) && (bool(entry.quality.inspected_first) || entry.quality.inspected_first === null)) line.quality = { checked_after_change: entry.quality.checked_after_change, fix_cycles_max: entry.quality.fix_cycles_max, inspected_first: entry.quality.inspected_first };
+  if (entry.quality && whole(entry.quality.fix_cycles_max) && (bool(entry.quality.checked_after_change) || entry.quality.checked_after_change === null) && (bool(entry.quality.inspected_first) || entry.quality.inspected_first === null)) {
+    line.quality = { checked_after_change: entry.quality.checked_after_change, fix_cycles_max: entry.quality.fix_cycles_max, inspected_first: entry.quality.inspected_first };
+    // 0.8.9: change blocks, how many a check followed, rework cycles.
+    if (whole(entry.quality.change_blocks) && whole(entry.quality.checked_blocks) && whole(entry.quality.rework_cycles) && entry.quality.checked_blocks <= entry.quality.change_blocks) Object.assign(line.quality, { change_blocks: entry.quality.change_blocks, checked_blocks: entry.quality.checked_blocks, rework_cycles: entry.quality.rework_cycles });
+  }
+  if (STAGES.includes(entry.stage)) line.stage = entry.stage;
+  // 0.9.0: how fast the stretch moved; whole seconds and counts, each key optional.
+  if (entry.timing && typeof entry.timing === "object") {
+    const timing = Object.fromEntries(TIMING_KEYS.filter((key) => whole(entry.timing[key])).map((key) => [key, entry.timing[key]]));
+    if (Object.keys(timing).length > 0) line.timing = timing;
+  }
+  // 0.9.1: risk and hygiene; whole counts, and whether a check followed delivery.
+  if (entry.hygiene && typeof entry.hygiene === "object") {
+    const hygiene = Object.fromEntries(HYGIENE_KEYS.filter((key) => whole(entry.hygiene[key])).map((key) => [key, entry.hygiene[key]]));
+    if (typeof entry.hygiene.checked_after_delivery === "boolean") hygiene.checked_after_delivery = entry.hygiene.checked_after_delivery;
+    if (Object.keys(hygiene).length > 0) line.hygiene = hygiene;
+  }
   const risk = pick(entry.risk, ["proposed", "refused", "run"]);
   if (risk && risk.proposed > 0 && risk.refused + risk.run <= risk.proposed) line.risk = risk;
   const keyed = (record, keys) => { const kept = Object.entries(record && typeof record === "object" ? record : {}).filter(([key, n]) => keys.includes(key) && whole(n) && n > 0).sort(); return kept.length ? Object.fromEntries(kept) : null; };
@@ -162,7 +178,13 @@ export function archiveEntries(dir, ownKey) {
  * a line of their own that names the stretch it adds to (`supplements`) and carries only what no earlier line of that
  * stretch carries, in the same chain, under the same day's root. Null when there is nothing new.
  */
-export const DERIVED_KEYS = ["signals", "analyzer_version", "verification", "unconfirmed", "delivery", "recovery", "delegation", "context", "routing", "steering", "tools", "complexity", "oversight", "planning", "changes", "autonomy", "authorship", "framing", "quality", "risk", "tool_mix", "reads", "context_files", "interrupts", "steers", "utc_offset"];
+/** The furthest step a stretch reached on the computer (0.8.9). */
+const STAGES = ["attempted", "verified", "committed", "pushed", "pr", "deployed"];
+/** How fast a stretch moved (0.9.0). */
+/** Risk and hygiene (0.9.1). */
+const HYGIENE_KEYS = ["rollbacks", "privileged", "bypasses", "secrets", "destructive", "destructive_checked", "guard_retried", "guard_changed", "guard_stopped"];
+const TIMING_KEYS = ["first_action_s", "first_check_s", "detect_actions", "detect_s", "recovery_calls", "resume_actions"];
+export const DERIVED_KEYS = ["signals", "analyzer_version", "verification", "unconfirmed", "delivery", "recovery", "delegation", "context", "routing", "steering", "tools", "complexity", "oversight", "planning", "changes", "autonomy", "authorship", "framing", "quality", "risk", "tool_mix", "reads", "context_files", "stage", "timing", "hygiene", "interrupts", "steers", "utc_offset"];
 export function supplementFor(line, earlier) {
   const missing = DERIVED_KEYS.filter((key) => line[key] !== undefined && !earlier.some((old) => old[key] !== undefined));
   if (missing.length === 0) return null;

package/stretch-evidence.mjs

@@ -36,6 +36,29 @@ export const callKindsOf = (block) => {
   const command = block && SHELL_TOOLS.has(String(block.name)) ? block.input?.command ?? block.input?.CommandLine ?? block.input?.cmd : null;
   return typeof command === "string" ? CALL_KINDS.filter(([, pattern]) => pattern.test(command)).map(([kind]) => kind) : [];
 };
+/**
+ * RISK AND HYGIENE MARKS (0.9.1), beside the kinds and outside a call's family so no earlier measure moves: a ROLLBACK
+ * (git revert or restore, a hard reset, a deploy rolled back), a PRIVILEGED command (sudo, a forced push, a world-
+ * writable chmod), a BYPASS (hooks skipped with --no-verify), and a SECRET touched (a .env, a private key, a credentials
+ * or token file read or printed). The path and the command are read here and never kept; only the mark travels.
+ */
+const MARKS = [
+  ["rollback", /\bgit\s+(revert|restore)\b|\bgit\s+reset\s+--hard\b|\b(vercel|fly|netlify)\s+rollback\b|\bkubectl\s+rollout\s+undo\b/],
+  ["privilege", /(^|[\s;&|])sudo\s|\bgit\s+push\b[^\n]*\s(-f|--force(-with-lease)?)\b|\bchmod\s+(-R\s+)?777\b/],
+  ["bypass", /\s--no-verify\b/],
+];
+const SECRET_FILE = /(^|[\/\s'"])(\.env(\.[\w-]+)?|id_(rsa|ed25519|ecdsa)|[\w.-]+\.(pem|p12|pfx|key)|credentials(\.json)?|\.npmrc|\.netrc|\.pgpass)(['"\s]|$)/;
+const PRINTS = /\b(cat|less|more|head|tail|bat|type|Get-Content|grep|sed|awk|cp|scp|base64)\b/;
+const READ_TOOLS = new Set(["Read", "read_file", "view", "open_file"]);
+export const marksOf = (block) => {
+  const name = String(block?.name ?? "");
+  const command = SHELL_TOOLS.has(name) ? block.input?.command ?? block.input?.CommandLine ?? block.input?.cmd : null;
+  const path = READ_TOOLS.has(name) ? block.input?.file_path ?? block.input?.path ?? null : null;
+  const marks = typeof command === "string" ? MARKS.filter(([, pattern]) => pattern.test(command)).map(([mark]) => mark) : [];
+  if ((typeof path === "string" && SECRET_FILE.test(path)) || (typeof command === "string" && PRINTS.test(command) && SECRET_FILE.test(command))) marks.push("secret");
+  return marks;
+};
+
 /**
  * A TOOL RESULT AS THE STRETCH READS IT (0.8.1): its call's id, whether it failed, and, read here and never kept, WHO
  * refused it when it was refused: the PERSON (Claude Code's fixed sentence when they decline a tool use or a plan) or a
@@ -54,7 +77,7 @@ export const resultOf = (block) => {
 const planOf = (block) => (block.name === "TodoWrite" && Array.isArray(block.input?.todos) ? { items: block.input.todos.length, done: block.input.todos.filter((todo) => todo && todo.status === "completed").length } : null);
 
 /** One tool call as the stretch reads it: its id, its kinds, its FAMILY (the tool and its check kinds) and a digest of its input (compared, never kept). */
-export const callOf = (block) => { const kinds = callKindsOf(block), plan = planOf(block); return { id: block.id, kinds, family: `${String(block.name)}|${kinds.join("+")}`, digest: createHash("sha256").update(JSON.stringify(block.input ?? null)).digest("base64url").slice(0, 16), ...(plan ? { plan } : {}) }; };
+export const callOf = (block) => { const kinds = callKindsOf(block), plan = planOf(block), marks = marksOf(block); return { id: block.id, kinds, ...(marks.length ? { marks } : {}), family: `${String(block.name)}|${kinds.join("+")}`, digest: createHash("sha256").update(JSON.stringify(block.input ?? null)).digest("base64url").slice(0, 16), ...(plan ? { plan } : {}) }; };
 
 /**
  * VERIFICATION PER STRETCH (0.7.1; framework L7, L12): per kind of check, how many ran and how many failed; per step of
@@ -268,6 +291,10 @@ function autonomyOf(messages) {
  *            (plan mode, a plan approved, a to-do list) came before it.
  *   quality: whether a check passed AFTER the stretch's last change; the most failures one check family took before it
  *            passed; whether the change was looked at (git diff, show, status) before the first delivery step.
+ *            0.8.9: change BLOCKS (edits with no check between them) and how many a check followed before the next one;
+ *            REWORK cycles (a check failed, then a change before that check ran again: change, fail, change).
+ *   stage:   the furthest step the stretch reached on this computer: attempted (it changed something), verified (a check
+ *            passed), committed, pushed, pr, deployed (0.8.9). Merged and confirmed are the server's, from GitHub.
  *   risk:    destructive commands proposed, refused (by the person or a guard) and run.
  *   tool_mix: calls per kind of tool (search, read, edit, shell, agent, web, data, mcp, plan).
  *   reads:   files read again unchanged (the same read twice), the cost of context that did not hold.
@@ -282,6 +309,8 @@ function practiceOf(messages, { isHumanTurn }) {
   let turns = 0, acted = false, planned = false, turnsBefore = 0, plannedFirst = false;
   let lastChange = -1, passedAfter = false, delivered = false, inspectedFirst = false;
   const failuresBefore = new Map(); let cyclesMax = 0;
+  let blocks = 0, checkedBlocks = 0, inBlock = false, rework = 0, passedAny = false;
+  const pendingFail = new Set(), reached = new Set();
   const risk = { proposed: 0, refused: 0, run: 0 }, mix = {}, reads = new Map();
   let index = 0;
   for (const message of messages) {
@@ -294,15 +323,21 @@ function practiceOf(messages, { isHumanTurn }) {
       mix[toolKind(tool)] = (mix[toolKind(tool)] ?? 0) + 1;
       if (PLAN_TOOLS.has(tool) && tool !== "AskUserQuestion") planned = true;
       if (!acted && ACTION_TOOLS.has(tool)) { acted = true; turnsBefore = turns; plannedFirst = planned; }
-      if (EDIT_TOOLS.has(tool)) { lastChange = index; passedAfter = false; }
+      if (EDIT_TOOLS.has(tool)) {
+        lastChange = index; passedAfter = false;
+        if (!inBlock) { blocks += 1; inBlock = true; }
+        rework += pendingFail.size; pendingFail.clear();
+      }
       if (tool === "Read") reads.set(call.digest, (reads.get(call.digest) ?? 0) + 1);
       const checks = call.kinds.filter((kind) => CHECK_KINDS.includes(kind));
+      if (checks.length > 0 && inBlock) { checkedBlocks += 1; inBlock = false; }
       for (const kind of checks) {
-        if (result?.failed === true) failuresBefore.set(kind, (failuresBefore.get(kind) ?? 0) + 1);
+        if (result?.failed === true) { failuresBefore.set(kind, (failuresBefore.get(kind) ?? 0) + 1); pendingFail.add(kind); }
+        if (result?.failed === false) { passedAny = true; pendingFail.delete(kind); }
         if (result?.failed === false) { cyclesMax = Math.max(cyclesMax, failuresBefore.get(kind) ?? 0); failuresBefore.set(kind, 0); if (index > lastChange) passedAfter = true; }
       }
       if (call.kinds.includes("inspect") && !delivered) inspectedFirst = true;
-      if (call.kinds.some((kind) => DELIVERY_KINDS.includes(kind)) && result?.failed === false) delivered = true;
+      if (call.kinds.some((kind) => DELIVERY_KINDS.includes(kind)) && result?.failed === false) { delivered = true; for (const kind of call.kinds) if (DELIVERY_KINDS.includes(kind)) reached.add(kind); }
       if (call.kinds.includes("destructive")) {
         risk.proposed += 1;
         if (result?.refusal) risk.refused += 1; else if (result?.failed === false) risk.run += 1;
@@ -311,18 +346,100 @@ function practiceOf(messages, { isHumanTurn }) {
   }
   if (index === 0) return null;
   const repeated = [...reads.values()].reduce((sum, n) => sum + Math.max(0, n - 1), 0);
+  const stage = ["deploy", "pr", "push", "commit"].find((kind) => reached.has(kind)) ?? (lastChange > 0 ? (passedAny ? "verified" : "attempted") : null);
   return {
     framing: { turns_before_action: acted ? turnsBefore : turns, planned_first: acted ? plannedFirst : planned },
-    quality: { checked_after_change: lastChange > 0 ? passedAfter : null, fix_cycles_max: cyclesMax, inspected_first: delivered ? inspectedFirst : null },
+    quality: { checked_after_change: lastChange > 0 ? passedAfter : null, fix_cycles_max: cyclesMax, inspected_first: delivered ? inspectedFirst : null, change_blocks: blocks, checked_blocks: checkedBlocks, rework_cycles: rework },
+    ...(stage ? { stage: { deploy: "deployed", pr: "pr", push: "pushed", commit: "committed" }[stage] ?? stage } : {}),
     ...(risk.proposed > 0 ? { risk } : {}),
     tool_mix: mix,
     reads: { repeated },
   };
 }
 
+/**
+ * HOW FAST A STRETCH MOVED (0.9.0): seconds from the person's first turn to the agent's first action (an edit, a write or
+ * a shell command) and from the first change to the first check; per failure, the actions and seconds until the next
+ * change that answered it (detection), and the calls from a failed check to that check passing (recovery depth); after
+ * a resume, a clear or a compaction, the calls until the next change or check (back to work). Medians, from the order
+ * and the clocks alone; nothing of what was said or run.
+ */
+const median = (values) => { if (values.length === 0) return null; const sorted = [...values].sort((a, b) => a - b); const middle = Math.floor(sorted.length / 2); return Math.round(sorted.length % 2 ? sorted[middle] : (sorted[middle - 1] + sorted[middle]) / 2); };
+function timingOf(messages, { isHumanTurn, textOf }) {
+  let firstTurn = null, firstAction = null, firstChange = null, firstCheck = null, index = 0, resumedAt = null;
+  const kindsById = new Map(), open = [], openChecks = new Map(), detect = [], detectSeconds = [], depth = [], resume = [];
+  for (const message of messages) {
+    if (message.bridge || !Number.isFinite(message.at)) continue;
+    const text = message.type === "user" ? textOf(message.content) : "";
+    if (message.compacted || ["compact", "clear", "resume"].includes(COMMAND.exec(text)?.[1] ?? "")) resumedAt = index;
+    if (message.type === "user" && !message.meta && message.kind !== "tool_result" && isHumanTurn(message.content) && firstTurn === null) firstTurn = message.at;
+    for (const result of message.results ?? []) {
+      const kinds = kindsById.get(result.id) ?? [];
+      if (result.failed !== true) {
+        for (const kind of kinds.filter((item) => CHECK_KINDS.includes(item) && result.failed === false)) { const at = openChecks.get(kind); if (at !== undefined) { depth.push(index - at); openChecks.delete(kind); } }
+        continue;
+      }
+      open.push({ index, at: message.at });
+      for (const kind of kinds.filter((item) => CHECK_KINDS.includes(item))) if (!openChecks.has(kind)) openChecks.set(kind, index);
+    }
+    for (const call of message.calls ?? []) {
+      index += 1;
+      kindsById.set(call.id, call.kinds);
+      const tool = call.family.split("|")[0];
+      if (firstAction === null && ACTION_TOOLS.has(tool)) firstAction = message.at;
+      if (EDIT_TOOLS.has(tool)) {
+        if (firstChange === null) firstChange = message.at;
+        for (const failure of open.splice(0)) { detect.push(index - failure.index); detectSeconds.push(Math.max(0, (message.at - failure.at) / 1000)); }
+      }
+      if (firstChange !== null && firstCheck === null && call.kinds.some((kind) => CHECK_KINDS.includes(kind))) firstCheck = message.at;
+      if (resumedAt !== null && (EDIT_TOOLS.has(tool) || call.kinds.some((kind) => CHECK_KINDS.includes(kind)))) { resume.push(index - resumedAt); resumedAt = null; }
+    }
+  }
+  const out = {
+    first_action_s: firstTurn !== null && firstAction !== null && firstAction >= firstTurn ? Math.round((firstAction - firstTurn) / 1000) : null,
+    first_check_s: firstChange !== null && firstCheck !== null ? Math.round((firstCheck - firstChange) / 1000) : null,
+    detect_actions: median(detect), detect_s: median(detectSeconds), recovery_calls: median(depth), resume_actions: median(resume),
+  };
+  const kept = Object.fromEntries(Object.entries(out).filter(([, value]) => value !== null));
+  return Object.keys(kept).length > 0 ? kept : null;
+}
+
+/**
+ * RISK AND HYGIENE PER STRETCH (0.9.1): rollbacks, privileged commands, hooks skipped, secrets touched; destructive
+ * commands and how many a look (git diff, show, status) or a check came before since the last change; after a guard
+ * refused a call, whether the next call was the same one again (retried), another (changed), or none (stopped); and
+ * whether a check ran after the stretch's first delivery step (a push, a deploy). Counts only.
+ */
+function hygieneOf(messages) {
+  const results = new Map();
+  for (const message of messages) for (const result of message.results ?? []) results.set(result.id, result);
+  const out = { rollbacks: 0, privileged: 0, bypasses: 0, secrets: 0, destructive: 0, destructive_checked: 0, guard_retried: 0, guard_changed: 0, guard_stopped: 0 };
+  let lookedSinceChange = false, delivered = false, checkedAfterDelivery = false, refusedDigest = null;
+  for (const message of messages) {
+    if (message.bridge) continue;
+    for (const call of message.calls ?? []) {
+      const tool = call.family.split("|")[0], result = results.get(call.id), marks = call.marks ?? [];
+      if (refusedDigest !== null) { if (call.digest === refusedDigest) out.guard_retried += 1; else out.guard_changed += 1; refusedDigest = null; }
+      if (marks.includes("rollback")) out.rollbacks += 1;
+      if (marks.includes("privilege")) out.privileged += 1;
+      if (marks.includes("bypass")) out.bypasses += 1;
+      if (marks.includes("secret")) out.secrets += 1;
+      if (call.kinds.includes("destructive")) { out.destructive += 1; if (lookedSinceChange) out.destructive_checked += 1; }
+      if (EDIT_TOOLS.has(tool)) lookedSinceChange = false;
+      if (call.kinds.includes("inspect") || call.kinds.some((kind) => CHECK_KINDS.includes(kind))) { lookedSinceChange = true; if (delivered) checkedAfterDelivery = true; }
+      if ((call.kinds.includes("push") || call.kinds.includes("deploy")) && result?.failed === false) delivered = true;
+      if (result?.refusal === "guard") refusedDigest = call.digest;
+    }
+  }
+  if (refusedDigest !== null) out.guard_stopped += 1;
+  const kept = Object.fromEntries(Object.entries(out).filter(([, value]) => value > 0));
+  if (delivered) kept.checked_after_delivery = checkedAfterDelivery;
+  return Object.keys(kept).length > 0 ? kept : null;
+}
+
 /** The stretch's whole derived record; `helpers` are the hook's own readers of a human turn, so both read it the same way. */
 export function deriveStretch(messages, helpers) {
   const context = contextOf(messages, helpers), routing = routingOf(messages), steering = steeringOf(messages, helpers);
-  const tools = toolsOf(messages), oversight = oversightOf(messages), planning = planningOf(messages), autonomy = autonomyOf(messages), practice = practiceOf(messages, helpers);
-  return { ...(practice ?? {}), ...(oversight ? { oversight } : {}), ...(planning ? { planning } : {}), ...(autonomy ? { autonomy } : {}), ...verificationOf(messages, helpers), ...(context ? { context } : {}), ...(routing ? { routing } : {}), ...(steering ? { steering } : {}), ...(tools ? { tools } : {}) };
+  const timing = timingOf(messages, helpers), hygiene = hygieneOf(messages), tools = toolsOf(messages), oversight = oversightOf(messages), planning = planningOf(messages), autonomy = autonomyOf(messages), practice = practiceOf(messages, helpers);
+  return { ...(practice ?? {}), ...(oversight ? { oversight } : {}), ...(planning ? { planning } : {}), ...(autonomy ? { autonomy } : {}), ...verificationOf(messages, helpers), ...(context ? { context } : {}), ...(routing ? { routing } : {}), ...(steering ? { steering } : {}), ...(tools ? { tools } : {}), ...(timing ? { timing } : {}), ...(hygiene ? { hygiene } : {}) };
 }

package/worktrust.mjs

@@ -73,7 +73,7 @@ const command = args.find((arg, at) => !arg.startsWith("--") && !(at > 0 && VALU
 const flag = (name) => { const at = args.indexOf(`--${name}`); return at >= 0 ? args[at + 1] : undefined; };
 const has = (name) => args.includes(`--${name}`);
 /** This CLI's version, said to the door so the app can tell which computer runs an old one (check-cli-package holds it equal to package.json). */
-const CLI_VERSION = "0.8.8";
+const CLI_VERSION = "0.9.1";
 const ORIGIN = (flag("origin") ?? process.env.WORKTRUST_ORIGIN ?? "https://app.worktrust.io").replace(/\/$/, "");
 const MCP = flag("url") ?? process.env.WORKTRUST_MCP_URL ?? `${ORIGIN}/api/mcp`;
 const HOME_DIR = join(homedir(), ".worktrust");