@event4u/agent-config 9.33.0 → 9.34.0

package/dist/agent-src/skills/skill-writing/SKILL.md

@@ -530,7 +530,7 @@ tool skill*, **never** a new skill per pitfall and never a generated grid (see
 [`size-enforcement`](../../rules/size-enforcement.md) § Per-tool pitfall
 content).
 
-## Rationalizations-to-reject section (required, security-stop-routed skills)
+## Rationalizations-to-reject section (recommended pattern, security-stop-routed skills)
 
 A skill routed by `security-sensitive-stop` already carries a forbidden-moves or
 failure-mode list. Those describe **what is wrong**. They do not answer the
@@ -553,7 +553,7 @@ the skip actually happens.
 Keep it to the arguments that have really been made. An invented rationalization
 teaches nothing and dilutes the ones that recur.
 
-## Non-negotiable-deliverable section (required, adjacent-technology clusters)
+## Non-negotiable-deliverable section (recommended pattern, adjacent-technology clusters)
 
 Some skills sit next to a technology that is a *tempting wrong answer* — close
 enough to look interchangeable, different enough that picking it silently loses
@@ -644,6 +644,84 @@ This duplicates the rules on purpose, and the duplication is the point: the
 rules are the enforcement, the section is what survives the artifact leaving
 their reach.
 
+## Mechanism-teaching section (optional pattern, skills whose subject has one)
+
+A procedure tells the agent what to do in the cases you thought of. A
+**mechanism** tells it why those steps work, which is the only thing that
+transfers to the case you did not. Where a skill's subject has a mechanism worth
+one screen — how the cascade resolves, why the queue drops the message, what the
+scheduler actually orders — state it in a `## How it works` section, ahead of the
+procedure that depends on it.
+
+Keep it to one screen and to mechanisms only: no history, no alternatives-considered,
+no restating the procedure in prose. If it takes more than a screen, the skill is
+carrying a guideline's worth of material and should point at one instead.
+
+The test for whether it earns its place: name a plausible situation the procedure
+does not cover, and check whether this section is enough to answer it. If the
+answer is still "read the source", the section is decoration.
+
+Optional, and it gets no gate: "teaches a mechanism" is a prose judgement, and a
+predicate over prose is the false-positive class this file's other patterns avoid.
+
+## Illustrative-not-verbatim marker (optional pattern, reference code)
+
+A code block in a skill is read as copy-me by default. When a block is teaching a
+*shape* — a signature to mirror, a structure to follow, pseudocode standing in for
+the real call — say so on the line above it:
+
+```
+<!-- illustrative: shape only, not a working snippet -->
+```
+
+One line, immediately above the fence, in the skill's own words if you prefer.
+What matters is that the reader can tell shape-teaching code from code that runs,
+because the failure mode is silent: an agent pastes the illustration, it almost
+works, and the debugging starts from the wrong premise.
+
+The inverse deserves the same care — a block that IS meant to be run verbatim,
+where a reader might assume otherwise, says so too. See also `### 4. Add
+safe/unsafe example`, which governs a different axis: safe-vs-unsafe, not
+runnable-vs-illustrative.
+
+## Headline metric + closing report (optional pattern, optimization skills)
+
+A skill whose job is to make something better — smaller, faster, cheaper, quieter
+— has to say **which single number it moves**. Without one, "optimized" is a
+claim nobody can check and every run reports success.
+
+So an optimization skill names, up front:
+
+1. **The headline metric.** One number, with its unit and how it is measured.
+   "Initial-context tokens, measured by `audit-tokens`" — not "context cost".
+2. **The closing report shape.** Before, after, delta, and the command that
+   produced both. The same command for both readings, or the delta is not a delta.
+
+State what the metric does **not** capture, in one line. An optimization that
+moves the headline number by degrading something it does not measure is the
+failure this pattern exists to make visible, and only the author knows where that
+edge is.
+
+## Contrastive-example slot (optional pattern, authoring and behaviour skills)
+
+Where a skill's guidance is easy to agree with and hard to apply — tone, phrasing,
+question shape, what counts as evidence — a wrong/right pair does what prose
+cannot: it shows the near-miss. Six live corpora already carry them:
+[`direct-answers-demos`](../../../docs/guidelines/agent-infra/direct-answers-demos.md),
+[`asking-and-brevity-examples`](../../../docs/guidelines/agent-infra/asking-and-brevity-examples.md),
+[`language-and-tone-examples`](../../../docs/guidelines/agent-infra/language-and-tone-examples.md),
+and `autonomy-examples` / `interrupt-examples` / `cheap-question-mechanics` under
+`src/agent-src/contexts/execution/`.
+
+Follow one of those rather than inventing a format. The shape they share: the
+**wrong** version first, in the form it actually gets written; the **right**
+version second; and one line of **why** — which is the load-bearing part, because
+a pair without it teaches the specific case and not the rule.
+
+Put the corpus in a guideline or context file and point at it from the skill when
+it outgrows a section. A near-miss corpus is reference material read on demand,
+not always-loaded prose.
+
 ## Do NOT
 
 * Write documentation-style, pointer-only, or too-broad skills ("Laravel skill", "Django skill")

package/dist/agent-src/skills/systematic-debugging/SKILL.md

@@ -222,10 +222,12 @@ right line beats five minutes of IDE breakpoints.
 
 **Known failure signature?** If the symptom has a recognisable shape — a
 tool/agent loop, HTTP 429, `ECONNREFUSED`, `ENOENT`, timeout/hang, OOM, flaky
-test, or works-locally-fails-in-CI — consult the symptom → cause → first-check
+test, works-locally-fails-in-CI, or repeated attempts at a tool that is not
+available — consult the symptom → cause → first-check
 lookup in [`failure-signatures`](../../../docs/guidelines/agent-infra/failure-signatures.md)
-to shortcut Phase 2 (Isolate) with the highest-probability first check. It
-supplements the loop; it does not replace it.
+to shortcut Phase 2 (Isolate) with the highest-probability first check. Each row
+carries a stable id you can cite, and a discrimination drill that produces its
+documented symptom. It supplements the loop; it does not replace it.
 
 ## Condition-based waiting (intermittent bugs)
 

package/dist/agent-src/skills/tailwind-engineer/SKILL.md

@@ -157,8 +157,32 @@ Risks:          <arbitrary values, !important, dark-mode gaps>
 
 When `DESIGN.md` declares `## Taste Dials`, honour them: Variance → layout-family spread + asymmetry tolerance; Motion → animation budget + reduced-motion posture; Density → spacing scale + information-per-viewport. Absent → follow the design brief's inferred dials.
 
+## Security constraints
+
+`scripts/tailwind_config_gen.ts` is the only shipped script.
+
+- **What it may touch** — exactly one file: `--output PATH`, or
+  `tailwind.config.ts` / `.js` in the current working directory when no
+  `--output` is given. It reads nothing from the project.
+- **What it must never do** — write anywhere other than that one path,
+  reach the network, or spawn a subprocess. It does none of these.
+- **Default invocation** — **mutating**, and stated here because it is a
+  real exception rather than a comfortable one: a bare invocation writes
+  `tailwind.config.ts` into `cwd`, overwriting an existing config without
+  prompting. `--validate-only` is the read-only path — it prints the
+  generated config to stdout and writes nothing. Run `--validate-only`
+  first, and pass `--output` when the target is not the cwd default.
+  This default contradicts `skill-writing`'s "never mutate on a bare
+  invocation" line; it is recorded rather than silently tolerated, and
+  changing the script's default is a caller-visible behaviour change that
+  belongs in its own change, not here.
+- **Outbound** — nothing. No network access.
+
 ## Do NOT
 
+- Do NOT run `tailwind_config_gen` bare in a project that already has a
+  `tailwind.config.*` — it overwrites without asking. Use
+  `--validate-only` first, then `--output` at the intended path.
 - Do NOT add `!important` to win a specificity fight; restructure
   the cascade or extract the conflicting style.
 - Do NOT introduce a new colour outside `tailwind.config` without

package/dist/agent-src/templates/agents/agent-project-settings.example.yml

@@ -39,7 +39,7 @@ schema_version: 1
 # CI guard: a release bump of `package.json` must update this value
 # in lockstep — see scripts/check_template_pin_drift.ts (road-to-
 # portable-runtime-and-update-check P3.3).
-agent_config_version: "9.33.0"
+agent_config_version: "9.34.0"
 
 # --- Project identity ---
 project:

package/dist/cli-delegate/{chunk-2WELHO3P.js → chunk-I5EAFRII.js}

@@ -449,6 +449,38 @@ function hasBoundaryMarker(block) {
   // outside the datamarking while still riding in the same block.
   block.trimEnd().endsWith(ENVELOPE_BOUNDARY_CLOSE);
 }
+var DECISION_REVERSIBILITY_TAGS = ["reversible", "irreversible"];
+function decisionTagErrors(decisions) {
+  if (!Array.isArray(decisions)) {
+    return [];
+  }
+  const errors = [];
+  for (const line of decisions) {
+    if (typeof line !== "string") {
+      continue;
+    }
+    const m = /\[([^\]]*)\]\s*$/.exec(line.trimEnd());
+    if (m === null) {
+      continue;
+    }
+    const inner = (m[1] ?? "").trim();
+    const lowered = inner.toLowerCase();
+    if (DECISION_REVERSIBILITY_TAGS.includes(lowered)) {
+      if (inner !== lowered) {
+        errors.push(
+          `decisions: reversibility tag must be lower-case \u2014 got \`[${inner}]\`, expected \`[${lowered}]\``
+        );
+      }
+      continue;
+    }
+    if (/^i?rr?ever[a-z]*[^a-z]*$/i.test(lowered)) {
+      errors.push(
+        `decisions: \`[${inner}]\` is not a reversibility tag \u2014 use exactly ` + DECISION_REVERSIBILITY_TAGS.map((t) => `\`[${t}]\``).join(" or ") + ", or drop the bracket"
+      );
+    }
+  }
+  return errors;
+}
 var RECYCLE_ENVELOPE_KEYS = /* @__PURE__ */ new Set([
   "capsule_version",
   "variant",
@@ -461,6 +493,7 @@ var RECYCLE_ENVELOPE_KEYS = /* @__PURE__ */ new Set([
   "not_carried_forward",
   "decisions",
   "constraints",
+  "do_not_touch",
   "open_worker_envelopes",
   "artifact_paths",
   "assumptions",
@@ -513,7 +546,9 @@ function validateRecycleEnvelope(input) {
   checkList(errors, "remaining", e["remaining"], MAX_LINE_CHARS, true);
   checkList(errors, "not_carried_forward", e["not_carried_forward"], MAX_LINE_CHARS, true);
   checkList(errors, "decisions", e["decisions"], MAX_LINE_CHARS, false);
+  errors.push(...decisionTagErrors(e["decisions"]));
   checkList(errors, "constraints", e["constraints"], MAX_LINE_CHARS, false);
+  checkList(errors, "do_not_touch", e["do_not_touch"], MAX_REF_CHARS, false);
   checkList(errors, "open_worker_envelopes", e["open_worker_envelopes"], MAX_REF_CHARS, false);
   checkList(errors, "artifact_paths", e["artifact_paths"], MAX_REF_CHARS, false);
   checkList(errors, "suggested_skills", e["suggested_skills"], MAX_REF_CHARS, false);

package/dist/cli-delegate/{chunk-PMQ6Z3T3.js → chunk-REMEZSVP.js}

@@ -1622,6 +1622,14 @@ import * as crypto from "node:crypto";
 import * as fs7 from "node:fs";
 import * as path10 from "node:path";
 import { fileURLToPath as fileURLToPath5 } from "node:url";
+
+// src/scripts/_lib/env_kill_switch.ts
+var _OFF_VALUES = /* @__PURE__ */ new Set(["", "0", "false", "False"]);
+function isEnvKillSwitchActive(name) {
+  return !_OFF_VALUES.has(process.env[name] ?? "");
+}
+
+// src/scripts/ai_council/events_log.ts
 var SCHEMA_VERSION = 2;
 var _VALID_ACTIONS = /* @__PURE__ */ new Set([
   "proceed",
@@ -1690,8 +1698,7 @@ function _utf8Bytes(cp) {
   );
 }
 function _kill_switch_active() {
-  const value = process.env[_KILL_SWITCH_ENV] ?? "";
-  return !(value === "" || value === "0" || value === "false" || value === "False");
+  return isEnvKillSwitchActive(_KILL_SWITCH_ENV);
 }
 function appendEvent(event, opts = {}) {
   if (_kill_switch_active()) {

package/dist/cli-delegate/{chunk-3EWYC3YK.js → chunk-UEZ7D35A.js}

@@ -271,7 +271,7 @@ function* iter_corpus(roots = DEFAULT_SCAN_ROOTS, exts = [".md"]) {
 function report(findings, opts) {
   const check_label = opts.check_label;
   if (findings.length === 0) {
-    process.stdout.write(`\u2705  ${check_label}: clean (${_corpus_note()}).
+    process.stdout.write(`\u2705  ${check_label}: clean (${_corpus_note(opts.scanned_roots)}).
 `);
     return 0;
   }
@@ -311,8 +311,9 @@ function _pyG(n) {
   }
   return s;
 }
-function _corpus_note() {
-  return "scanned " + DEFAULT_SCAN_ROOTS.join(", ");
+function _corpus_note(scanned_roots) {
+  const roots = scanned_roots && scanned_roots.length > 0 ? scanned_roots : DEFAULT_SCAN_ROOTS;
+  return "scanned " + roots.join(", ");
 }
 var PyFloat = class {
   constructor(value) {

package/dist/cli-delegate/cmd_conformance.js

@@ -19,7 +19,7 @@ import {
   frontmatterPaths,
   proseEqual,
   splitFrontmatter
-} from "./chunk-PMQ6Z3T3.js";
+} from "./chunk-REMEZSVP.js";
 import "./chunk-FTAISGDQ.js";
 import {
   atomic_write_json,

package/dist/cli-delegate/cmd_doctor.js

@@ -66,7 +66,7 @@ import {
   _wizard_state_path,
   main,
   shutilWhich
-} from "./chunk-PMQ6Z3T3.js";
+} from "./chunk-REMEZSVP.js";
 import "./chunk-FTAISGDQ.js";
 import "./chunk-EJITJBJX.js";
 import "./chunk-Q26HT4QT.js";

package/dist/cli-delegate/cmd_handoff.js

@@ -15,8 +15,8 @@ import {
   scanEnvelopeDirectives,
   validateRecycleEnvelope,
   wrapAsPriorSessionData
-} from "./chunk-2WELHO3P.js";
-import "./chunk-3EWYC3YK.js";
+} from "./chunk-I5EAFRII.js";
+import "./chunk-UEZ7D35A.js";
 import {
   readHookStdin
 } from "./chunk-MMJL4TIX.js";

package/dist/cli-delegate/cmd_memory_get.js

@@ -2,7 +2,7 @@
 import { createRequire as __acCreateRequire } from 'node:module'; const require = globalThis.require ?? __acCreateRequire(import.meta.url);
 import {
   memory_get_v1
-} from "./chunk-3EWYC3YK.js";
+} from "./chunk-UEZ7D35A.js";
 import "./chunk-OY53J6II.js";
 import "./chunk-J55WI3RR.js";
 

package/dist/cli-delegate/cmd_route_audit.js

@@ -3,7 +3,7 @@ import { createRequire as __acCreateRequire } from 'node:module'; const require
 import {
   read_entries
 } from "./chunk-QEEB4LHO.js";
-import "./chunk-3EWYC3YK.js";
+import "./chunk-UEZ7D35A.js";
 import "./chunk-MMJL4TIX.js";
 import {
   match_prompt

package/dist/cli-delegate/cmd_session_recycle.js

@@ -5,7 +5,7 @@ import {
   RECYCLE_ENVELOPE_REL,
   collectGrounding,
   validateRecycleEnvelope
-} from "./chunk-2WELHO3P.js";
+} from "./chunk-I5EAFRII.js";
 import {
   atomic_write_json
 } from "./chunk-EJITJBJX.js";
@@ -30,8 +30,13 @@ function templateEnvelope() {
     acceptance_criteria: ["<what done means, one line each>"],
     remaining: ["<open work, one line each>"],
     not_carried_forward: ["<what the successor must re-derive from source>"],
-    decisions: ["<decision \u2014 one-line rationale>"],
+    decisions: ["<decision \u2014 one-line rationale> [reversible]"],
     constraints: ["<binding constraint>"],
+    // Ships EMPTY, unlike the prose lists above: the entries are path refs,
+    // and a `<placeholder>` left in by a hurried writer would be a fake path
+    // rather than obvious filler. An empty list is also the honest default —
+    // nothing off limits is the common case.
+    do_not_touch: [],
     open_worker_envelopes: [],
     artifact_paths: ["<path to a deliverable / note / evidence file>"],
     assumptions: [],

package/dist/discovery/deprecation-report.md

@@ -1,6 +1,6 @@
 # Discovery — Deprecation Report
 
-- Generated: `2026-08-11T05:16:25Z`
+- Generated: `2026-08-11T20:26:09Z`
 - Deprecated artefacts: **0**
 
 _None. Tree is clean._

package/dist/discovery/discovery-manifest.json

@@ -2573,7 +2573,7 @@
     },
     {
       "category": "command",
-      "checksum": "sha256:b718faefe3c50937e0b4851d26086d9aa8a9573c74e3e9fd371d2d0935acaa02",
+      "checksum": "sha256:0181143b6ba9ee8ed18fc3069d175e1702a5a8b36bf9694fa4a07f7c98f52ca5",
       "install": {
         "default": true,
         "removable": false
@@ -9271,7 +9271,7 @@
     },
     {
       "category": "skill",
-      "checksum": "sha256:bae8a558172e50ef861a1e8e0132d24e2c52c4ac136b3c7562735b92c4d6d04d",
+      "checksum": "sha256:f7a67c0b8382e292ed0e98d36e683644c66b8291f404f75d321a89aed7574ae3",
       "install": {
         "default": true,
         "removable": true
@@ -9645,7 +9645,7 @@
     },
     {
       "category": "skill",
-      "checksum": "sha256:169fed9cc69e866501e9f8e4423aa7226cb92c0c0380a20a6f4d5a4f32ec6925",
+      "checksum": "sha256:7d2f4b3b162943305cf64778e2e63b67331f7861bf0e2eec392293e9579445d8",
       "install": {
         "default": true,
         "removable": true
@@ -12775,7 +12775,7 @@
     },
     {
       "category": "skill",
-      "checksum": "sha256:085ad155c77608b45ad63c5ebcfb3e08b2d7d6e69bd0d938860fe2f19eb245f0",
+      "checksum": "sha256:d547a77483a8022c0a61d7efcb254e287e966ff4bf942e0155ca41824859a130",
       "install": {
         "default": false,
         "removable": true
@@ -13237,7 +13237,7 @@
     },
     {
       "category": "skill",
-      "checksum": "sha256:671eb67f32df15e5a7a8e937ca694da71fd325a62e1303b57fdcc05899458514",
+      "checksum": "sha256:e0d38b712e69fa5f9bb8fe0ec9d8ba1020c2a45e0502e32a915259091e612a82",
       "install": {
         "default": true,
         "removable": false
@@ -13589,7 +13589,7 @@
     },
     {
       "category": "skill",
-      "checksum": "sha256:b2e8847e0acb03cad8c71f095d29d311ee76b4a11760ab9b21c5b07797b955df",
+      "checksum": "sha256:1a3c1965abf5ebfbeac3b17142c0bc84a955156ea8fa2928c048aeb540ce8ff1",
       "install": {
         "default": true,
         "removable": false
@@ -13809,7 +13809,7 @@
     },
     {
       "category": "skill",
-      "checksum": "sha256:c8b5af3f19e2c2de1b88a7ce140aa681f09006a2f44290303f709eb7bc1ba9cc",
+      "checksum": "sha256:1d7259677165ad794e4f16d652830d937ce21f56dbfc0aa440869163f3259792",
       "install": {
         "default": true,
         "removable": false
@@ -13831,7 +13831,7 @@
     },
     {
       "category": "skill",
-      "checksum": "sha256:47ca9fbe052b2046cc1897cf20cd52352de7ad557f7f1eed4d842f9291c53963",
+      "checksum": "sha256:89e099d1d527ed2000dd3bbe43dddac623487b6f75b08f4fb0a6b9c3ec39dfb8",
       "install": {
         "default": true,
         "removable": false
@@ -14486,7 +14486,7 @@
       "workspaces": []
     }
   ],
-  "checksum": "sha256:1a5894344548cfa9a3a79726d9063c47af89da73892e847856ca4748dad6550f",
+  "checksum": "sha256:578a1ff7f0b187812c772bce1b731880d7bca27d0f84a864a00b493e857fa21e",
   "deprecations": [
     {
       "key": "tier",
@@ -14629,7 +14629,7 @@
       "reason": "scaffold template for ticket-bundle tickets (ADR-101); carries no runtime frontmatter — emit-tickets fills it in per ticket"
     }
   ],
-  "generated_at": "2026-08-11T05:16:25Z",
+  "generated_at": "2026-08-11T20:26:09Z",
   "packs": [
     {
       "artefact_count": 118,
@@ -15482,7 +15482,7 @@
       ]
     }
   ],
-  "scanner_version": "c24312e116f8",
+  "scanner_version": "1eaf68b6e9c9",
   "stats": {
     "by_category": {
       "command": 196,

package/dist/discovery/discovery-manifest.json.sha256

@@ -1 +1 @@
-efd249af793682a1fa9969433c123c923e08e1eee8394865bb93a7c875c7cbec  discovery-manifest.json
+82fbc2b79bb7b97c663111542bbfdd08c3c798e17c6f992e19689a71c4d7403c  discovery-manifest.json

package/dist/discovery/discovery-manifest.summary.md

@@ -1,7 +1,7 @@
 # Discovery Manifest — Summary
 
-- Generated: `2026-08-11T05:16:25Z`
-- Scanner: `c24312e116f8`
+- Generated: `2026-08-11T20:26:09Z`
+- Scanner: `1eaf68b6e9c9`
 - Artefacts: **603**
 - Unassigned: **0**
 

package/dist/discovery/dormancy-report.md

@@ -0,0 +1,17 @@
+# Discovery — Dormancy Report
+
+- Generated: `2026-08-11T20:26:09Z`
+- Window: **6 months** without a commit
+
+> Dormancy triggers **review, never auto-deprecation**
+> (`docs/governance.md` § Skill lifecycle policy). A finished artefact and an
+> abandoned one look identical from commit dates, so this is a prompt to ask
+> "still earning its slot?" — never a removal list, and never a gate.
+
+## Signal unavailable
+
+**No dormancy list is published**: shallow clone — a truncated history is indistinguishable from a dormant artefact. Re-run after `git fetch --unshallow`.
+
+An empty section here would read as "nothing is dormant", which is a
+different claim from "the signal could not be computed".
+

package/dist/discovery/orphan-report.md

@@ -1,6 +1,6 @@
 # Discovery — Orphan Report
 
-- Generated: `2026-08-11T05:16:25Z`
+- Generated: `2026-08-11T20:26:09Z`
 - Orphan artefacts: **0**
 
 > An orphan is an artefact whose declared pack has no other members.

package/dist/discovery/packs.json

@@ -1,6 +1,6 @@
 {
-  "checksum": "sha256:1a5894344548cfa9a3a79726d9063c47af89da73892e847856ca4748dad6550f",
-  "generated_at": "2026-08-11T05:16:25Z",
+  "checksum": "sha256:578a1ff7f0b187812c772bce1b731880d7bca27d0f84a864a00b493e857fa21e",
+  "generated_at": "2026-08-11T20:26:09Z",
   "packs": [
     {
       "always_on": false,
@@ -1814,5 +1814,5 @@
       ]
     }
   ],
-  "scanner_version": "c24312e116f8"
+  "scanner_version": "1eaf68b6e9c9"
 }

package/dist/discovery/trust-report.md

@@ -1,6 +1,6 @@
 # Discovery — Trust Report
 
-- Generated: `2026-08-11T05:16:25Z`
+- Generated: `2026-08-11T20:26:09Z`
 - Workspaces tracked: **10**
 - Human-review-required artefacts: **3**
 

package/dist/discovery/workspaces.json

@@ -1,7 +1,7 @@
 {
-  "checksum": "sha256:1a5894344548cfa9a3a79726d9063c47af89da73892e847856ca4748dad6550f",
-  "generated_at": "2026-08-11T05:16:25Z",
-  "scanner_version": "c24312e116f8",
+  "checksum": "sha256:578a1ff7f0b187812c772bce1b731880d7bca27d0f84a864a00b493e857fa21e",
+  "generated_at": "2026-08-11T20:26:09Z",
+  "scanner_version": "1eaf68b6e9c9",
   "workspaces": [
     {
       "artefact_count": 202,