@event4u/agent-config 7.0.0 → 7.0.2

package/docs/archive/CHANGELOG-pre-7.0.0.md

@@ -0,0 +1,126 @@
+# Changelog Archive — pre-7.0.0
+
+> Frozen snapshot of `event4u/agent-config` changelog entries
+> released before `7.0.0`, split out of the main
+> [`CHANGELOG.md`](../../CHANGELOG.md) by `scripts/release.py`
+> once the active era's body crossed the drift cap enforced by
+> `tests/test_changelog_eras.py`.
+>
+> **Read-only.** New entries land in `CHANGELOG.md`. Entries
+> here are not amended — git tags remain the canonical source
+> for what shipped.
+>
+> Entry shape follows
+> [`../contracts/CHANGELOG-conventions.md`](../contracts/CHANGELOG-conventions.md).
+
+## [6.1.0](https://github.com/event4u-app/agent-config/compare/6.0.0...6.1.0) (2026-06-14)
+
+### Features
+
+* **rules:** Iron Law — never drop inherited/shared-branch commits without asking ([e03dc18](https://github.com/event4u-app/agent-config/commit/e03dc18300a2ef032d508062be032267a1f1f844))
+* **workspace:** plain-language host-decision explainability + employee-mode gate (6.0.0 P5) ([f485476](https://github.com/event4u-app/agent-config/commit/f485476050d3d7660aa3aaeacc54438e5eee44a9))
+* **governance:** workspace boundary contract + import-edge drift linter (6.0.0 P4) ([943eb29](https://github.com/event4u-app/agent-config/commit/943eb29747eb3dca0988317e2a526cd38d12f06c))
+* **install:** preview-only dry-run for the upgrade-cleanup reaper (6.0.0 P2) ([4caa802](https://github.com/event4u-app/agent-config/commit/4caa802b1fd1bbf1beda71e5ad553000e7b7f2bf))
+* **create-pr:** run the PR-gate archival sweep before opening the PR ([bdd645b](https://github.com/event4u-app/agent-config/commit/bdd645b3bcc7443c00df653d8f53982417d0ca3e))
+* **roadmaps:** deterministic PR-gate archival sweep ([10643ca](https://github.com/event4u-app/agent-config/commit/10643ca9f5ff56ff02f376531e5c8c07998aa08d))
+* **memory:** remove agent-memory Layer 2; file-first memory only ([c8903c6](https://github.com/event4u-app/agent-config/commit/c8903c6f3a5502434b29c1b55535a9ebc3c55ac4))
+* **bench:** v2 pilot run + honest-null report; spawn stratified follow-up (Phase 4-6) ([5b7f7bd](https://github.com/event4u-app/agent-config/commit/5b7f7bd1112462f83ca42ef6509f4871cdba7ebe))
+* **bench:** v2 dual-axis scorer + runner + paired stats (Phase 2-3) ([ae25ee2](https://github.com/event4u-app/agent-config/commit/ae25ee2c7a8b0fe6325a0c886911a8c48c7546d5))
+* **bench:** v2 discipline-axis corpus + fixtures (Phase 0-1) ([1371dd0](https://github.com/event4u-app/agent-config/commit/1371dd05096c38c7f28f0c4a55b1daf2bb230500))
+* **bench:** wrapper-lift A/B benchmark — model-pin, budget-cap, error-aware ([a09238a](https://github.com/event4u-app/agent-config/commit/a09238ad0455baaf6f68016a195ad54725acd3a3))
+* **reasoning:** add model-agnostic Reasoning Discipline Protocol (RDP) ([b4b5bec](https://github.com/event4u-app/agent-config/commit/b4b5bec18f58630add4e7802c9e32b46b88f3084))
+* **security:** Phase 3 consumer audit + runtime injection hook + review skill ([23efe50](https://github.com/event4u-app/agent-config/commit/23efe5001104c1fd7056f7b2216fc4583f7f29f3))
+* **security:** Phase 2 injection-aware authoring rules ([eb41da9](https://github.com/event4u-app/agent-config/commit/eb41da9c7a1e3140d1efad760bd06e0bd53cb513))
+* **security:** Phase 1 self-audit corpus linters + containment convention ([f678aba](https://github.com/event4u-app/agent-config/commit/f678abab7b521f389e16abb837594b3d23a38d5f))
+* **ai-council:** resolve config user-global-first ([9177b52](https://github.com/event4u-app/agent-config/commit/9177b52a99842b3adffc21b325b39610ef520c6d))
+* **ci:** guard against re-introducing external-source references ([88d35d9](https://github.com/event4u-app/agent-config/commit/88d35d913f5afec7db33f5817229b366cac7256f))
+* **doctor:** add read-only stale-orphans health check ([48e6118](https://github.com/event4u-app/agent-config/commit/48e6118eab2e39fbfa1884a9f6defb1c8719a2ba))
+* **packs:** split meta into capability-scoped packs ([b363839](https://github.com/event4u-app/agent-config/commit/b3638393a6c67ecaac41ae19df89b2ed0b03154b))
+* **commands:** add first-class visibility frontmatter field ([db4d88d](https://github.com/event4u-app/agent-config/commit/db4d88dcf41629f27973624d6476f935be32e2c9))
+
+### Bug Fixes
+
+* **refs:** mark the RDP-contract acceptance criterion as a forward-ref ([c7cee96](https://github.com/event4u-app/agent-config/commit/c7cee966495a5c96932c9f2b71ab72aae517623f))
+* **memory:** mine-session description ≤200 chars + command-count floor ([f368e06](https://github.com/event4u-app/agent-config/commit/f368e0667980e3f3a03233e8152fd23daaa8bc5d))
+* **bench:** lint-bench-ab validates the v2 docs/benchmark.md sections ([41d552b](https://github.com/event4u-app/agent-config/commit/41d552bd3a162577be3a898fca9123cce87b38b9))
+* **server:** mirror reasoning.* in settings schema (template parity + diff) ([5f77256](https://github.com/event4u-app/agent-config/commit/5f77256da1ecbfa03f3aa1261f23ec940991c2bc))
+* **reasoning:** skill-lint senior sections + valid rule tier (2b) ([7f3e278](https://github.com/event4u-app/agent-config/commit/7f3e2789ab483b6938a39aac303f1fbe2f35792e))
+* **security:** refresh threat-model command effective-hash ([5505b57](https://github.com/event4u-app/agent-config/commit/5505b57fa6e4e4bf02e2891f49a30e6dc9128da1))
+* **security:** tighten security-sensitive-stop — clear long_rule lint regression ([d528126](https://github.com/event4u-app/agent-config/commit/d528126120bf51335267bdde66793343ec376c87))
+* **security:** list agent-security-review + security-audit-config in marketplace.json ([f432a62](https://github.com/event4u-app/agent-config/commit/f432a62d42fab5fc6707b166e869ca8cc2d5d866))
+* **security:** CI — portability reword + settings-schema parity for injection_scan ([fa78a58](https://github.com/event4u-app/agent-config/commit/fa78a589101fd41984d85496666f2fd618e49a1f))
+* **cli:** preflight dependency guard with an actionable message ([ce58c76](https://github.com/event4u-app/agent-config/commit/ce58c7683541a3bc22852186a9764be127f87e4b))
+* **repo:** untrack stray node_modules symlink, ignore symlink form ([67cc9e6](https://github.com/event4u-app/agent-config/commit/67cc9e61deb241aa23e9affaa5adc1fd92d99f14))
+* **refs:** exempt gitignored agents/.harvest-local in check_references ([0d0b170](https://github.com/event4u-app/agent-config/commit/0d0b1702c27c06e96d486211f639148f7ed0628e))
+* **refs:** mark gitignored harvest-local ref in security-pillar roadmap ([3abf215](https://github.com/event4u-app/agent-config/commit/3abf215692cf49ac190248c0007668b5a7b1e1f0))
+* **ci:** guard against empty (0-byte) roadmap files ([5115865](https://github.com/event4u-app/agent-config/commit/511586591171735437bba8cdb106094628a3d611))
+* **roadmaps:** restore competitive-borrow v6 work, drop empty reaping stub ([9fceeb7](https://github.com/event4u-app/agent-config/commit/9fceeb7fc9585965f44e25543a9699d0ff8ee519))
+* **ai-council:** align user-global config path with the setup wizard ([54de5a8](https://github.com/event4u-app/agent-config/commit/54de5a853c271311d79b2b436dd3fa58fa25eb69))
+* **rule:** drop non-portable task-invocations from source-confidentiality backstop ([57e6f7f](https://github.com/event4u-app/agent-config/commit/57e6f7fe5199cc5192ead2a1468ad7d04dd5d5bc))
+* **install:** always-run tag sweep so reaping catches pre-inventory orphans ([91e5418](https://github.com/event4u-app/agent-config/commit/91e541852c30899cb84dd2070f705dfe7947d350))
+* **discovery:** register visibility + new capability packs in schema/vocab ([1fcc72b](https://github.com/event4u-app/agent-config/commit/1fcc72b08fbd13f85c01d3df7fbc98790395ac91))
+
+### Documentation
+
+* **domain-watch:** capture deferred heavyweight tracks in one watch-note (6.0.0 P6) ([e212e1f](https://github.com/event4u-app/agent-config/commit/e212e1fcebed5d61c63d0719deb485787aca6fde))
+* **changelog:** add 6.0.0 release overview + breaking-changes section (6.0.0 P3) ([fdb4b63](https://github.com/event4u-app/agent-config/commit/fdb4b63cd495489341215430ae12a9866cb8503b))
+* fix stale source-of-truth paths and drop above-fold jargon (6.0.0 P1) ([9ee9870](https://github.com/event4u-app/agent-config/commit/9ee98708e1fdcf612848b393aace48b33a548e99))
+* **roadmaps:** add capability-discoverability follow-up from fable-feedback-3 ([a658c2f](https://github.com/event4u-app/agent-config/commit/a658c2f9c338182d1db2db503ab8ce6aa9c1f42f))
+* **model-recommendation:** add orchestrator→subagent model-routing guidance ([669f83d](https://github.com/event4u-app/agent-config/commit/669f83de5b2e6de590d87f7a093eb098bcf8a086))
+* **roadmaps:** add RDP discoverability roadmap ([c65e1b2](https://github.com/event4u-app/agent-config/commit/c65e1b2648af8bfd2a39a6c6dc74124cfc67101b))
+* **roadmaps:** revise discipline-axis pilot for weak-host-first methodology ([9ccb816](https://github.com/event4u-app/agent-config/commit/9ccb816b1fe449cb537567696487c09a4ea2ee8e))
+* **rules:** replace Merge-gated with PR-gate in roadmap-progress-sync ([c140687](https://github.com/event4u-app/agent-config/commit/c1406876e880babd70a178bb958af89f31c49eec))
+* **roadmaps:** memory-layer-cleanup removal + consolidation plans ([254758f](https://github.com/event4u-app/agent-config/commit/254758f1a24d7898b5828a3396d20b7498370d08))
+* **roadmap:** check-refs:skip discipline-axis bench (external repo citations) ([13c5fdb](https://github.com/event4u-app/agent-config/commit/13c5fdb04016c40488140f59688e9db88eee0d9d))
+* **roadmaps:** agent-memory removal + memory-pipeline consolidation plans ([6fbf6a6](https://github.com/event4u-app/agent-config/commit/6fbf6a6d33571839f8c85b276a8728ff62c606c7))
+* **roadmaps:** 3-condition value bench (v1) + discipline-axis bench (v2) ([f5a89c6](https://github.com/event4u-app/agent-config/commit/f5a89c69e60c345dbb655f3b0dab294d2b56669a))
+* **roadmaps:** RDP eval+promotion follow-up; archive frontier-reasoning parent ([417223b](https://github.com/event4u-app/agent-config/commit/417223bce4a4d04404512cfa804ad872dd11324c))
+* **roadmap:** security pillar — supply-chain integrity + injection-aware authoring ([c15ef6a](https://github.com/event4u-app/agent-config/commit/c15ef6adc17226513660b324b7960dc80125bbf3))
+* **roadmap:** fill in road-to-6.0.0-final-readiness ([46ba5a0](https://github.com/event4u-app/agent-config/commit/46ba5a002fb300afbce43e56b92a945bd0957dec))
+* **roadmap:** add image/brand/typography + greenfield-scaffold roadmaps ([f615452](https://github.com/event4u-app/agent-config/commit/f6154528a7b1ba34fb5543c91ac2c905999d2e8c))
+* **ai-council:** repoint config references to the user-global path ([1307f80](https://github.com/event4u-app/agent-config/commit/1307f804b2c1cc4a2b9a3d2ad1b829f2a7108619))
+* **ai-council:** document user-global config location (ADR-093) ([c501e55](https://github.com/event4u-app/agent-config/commit/c501e557fdb94d7214555a3b53042a275b2dc923))
+* **roadmap:** competitive-borrow plan (source-anonymous) ([9e3c692](https://github.com/event4u-app/agent-config/commit/9e3c692745170ad62b0f67d40afa8aad751e34dd))
+* **roadmap:** close metadata-leanness, spawn tier-removal follow-up ([93d5361](https://github.com/event4u-app/agent-config/commit/93d5361d05d5ddef944d4931c6260e69464d8270))
+* **adr:** ADR-092 defer command tier-alias removal (forcing-function) ([71d2630](https://github.com/event4u-app/agent-config/commit/71d26300d14c9b894dfef19ed6a6b3c996f5895f))
+* **roadmap:** metadata & command-surface leanness plan + evidence ([b31d106](https://github.com/event4u-app/agent-config/commit/b31d1061f5816eb3a8aadd4bd154e76ee00d089a))
+* **roadmap:** archive completed reaping-orphans fix plan ([c1ed447](https://github.com/event4u-app/agent-config/commit/c1ed4478f14c634c5fde5dbed224f72ae2c6eedb))
+* record evidence-gated decisions, archive residuals roadmap ([3657e12](https://github.com/event4u-app/agent-config/commit/3657e12fbcacb4740b4d1d84309d7dea28cd3c58))
+
+### Refactoring
+
+* **ai-council:** relocate config out of the repo to user-global ([17fb14a](https://github.com/event4u-app/agent-config/commit/17fb14a03736cb14f2f56861f194e5bfc00e80b9))
+* **commands:** fold fix:pr-bot/developer-comments into fix:pr-comments ([4c1c1c8](https://github.com/event4u-app/agent-config/commit/4c1c1c8795d422e1ba8d5ae77409f43f3b8fb2e5))
+
+### Tests
+
+* lower command-surface floor to 148 after fix:pr-* fold ([2b5e9ef](https://github.com/event4u-app/agent-config/commit/2b5e9ef7f6a8160aac6d2d26bb2876d7d8ed1427))
+* **install:** lock the reaping-gap class against regression ([d7ca6d0](https://github.com/event4u-app/agent-config/commit/d7ca6d0f967672d02d0b367fcfa408f6d675a220))
+
+### CI
+
+* **skill-lint:** also enforce rule tiers + guard the CI wiring ([6cde280](https://github.com/event4u-app/agent-config/commit/6cde2809c5ace248807236dee9634a590955795d))
+* **skill-lint:** run lint-command-tiers on every PR ([2e123c6](https://github.com/event4u-app/agent-config/commit/2e123c696a6bf2047cbc4473a9555c51b8f33da8))
+
+### Chores
+
+* **roadmaps:** complete + archive road-to-6.0.0-final-readiness (6.0.0) ([652294a](https://github.com/event4u-app/agent-config/commit/652294ad01ad48a2875ebd75953b82d06c350e56))
+* **roadmaps:** archive completed memory-layer-cleanup roadmaps ([9380482](https://github.com/event4u-app/agent-config/commit/9380482cf512c73f0783ff7a2270793d5dca8b95))
+* **memory:** regenerate pack manifest for shortened mine-session description ([14cec0e](https://github.com/event4u-app/agent-config/commit/14cec0ed0c6de254639e3727d7ad1f3a3147902e))
+* **memory:** regenerate projections, catalogs, and dashboard ([78cc1a1](https://github.com/event4u-app/agent-config/commit/78cc1a1180f6eced4c60e0c408da1103d0621ab0))
+* **roadmaps:** regenerate progress dashboard ([d0d2757](https://github.com/event4u-app/agent-config/commit/d0d2757e93026fab6f995732a130f126371e7978))
+* **security:** archive road-to-security-pillar (Phases 1-3 shipped) ([ad5f552](https://github.com/event4u-app/agent-config/commit/ad5f55204bae429b960db40976c3dc40ea759f97))
+* **security:** bump command-count messaging 148→149 (new /security-audit-config) ([5f5b79a](https://github.com/event4u-app/agent-config/commit/5f5b79a1d741b1565cfbc0aaf8c74aa43165fd0b))
+* **security:** regenerate dist, router, counts, pack manifests ([84ff0fc](https://github.com/event4u-app/agent-config/commit/84ff0fcd1216fbbd05a3a340148f681b95c2d4a8))
+* add uncomitted roadmaps ([699c2c5](https://github.com/event4u-app/agent-config/commit/699c2c5e8a70453a46b9c70e68ba621f96a5f64c))
+* **condense:** reconcile stale condensation hashes ([d9ac27e](https://github.com/event4u-app/agent-config/commit/d9ac27e232cc60dba941f75632dde9a5bbd2c4db))
+* add uncomitted roadmaps ([28ff06c](https://github.com/event4u-app/agent-config/commit/28ff06ce3741db46a5d39e1bc0a2a698d3a94c1f))
+* **rules:** track anonymized harvest roadmaps; harden source-confidentiality ([bc77fb9](https://github.com/event4u-app/agent-config/commit/bc77fb93476c8045cd227f94ad09b0f1f3cebbdd))
+* sync stale condensation for work / agents-init / agents-optimize ([0f816ca](https://github.com/event4u-app/agent-config/commit/0f816cae017824af718cd750457bfc1476252927))
+* sync command-count messaging to 148 after fix:pr-* removal ([94097ce](https://github.com/event4u-app/agent-config/commit/94097ce0bee1739d2f54688705b1aaf78a73cee4))
+* sync stale condensation hashes for unchanged command files ([a32da9a](https://github.com/event4u-app/agent-config/commit/a32da9a5c4199d1148549f63bf7c0012496cd2b6))
+* remove external-source references from the tracked tree ([3248611](https://github.com/event4u-app/agent-config/commit/3248611a8ae2e7d0574ba8e121f85077faae9f69))
+* **commands:** backfill visibility + re-tag pack across command sources ([d08f5b6](https://github.com/event4u-app/agent-config/commit/d08f5b6550735db22f4f855f155d2fd91a8dc047))
+* **changelog:** split era 5.9.x → pre-6.0.0 ([c442750](https://github.com/event4u-app/agent-config/commit/c44275053c07352402e162d6accbe67f6d380010))
+
+Tests: 6068

package/package.json

@@ -1,6 +1,6 @@
 {
     "name": "@event4u/agent-config",
-    "version": "7.0.0",
+    "version": "7.0.2",
     "description": "Universal AI Agent OS — audited skills, governance rules, commands, and templates for AI coding tools (Claude Code, Cursor, Windsurf, Copilot).",
     "license": "MIT",
     "private": false,

package/src/scripts/_lib/changelog_eras.ts

@@ -126,6 +126,64 @@ export function current_era_body_size(lines: readonly string[] | null = null): n
     return next_era_line - current_idx - 1;
 }
 
+/**
+ * Lines in the current era body EXCLUDING the newest (topmost) released
+ * version section.
+ *
+ * The drift cap bounds *accumulated prior* releases in the current era — the
+ * mechanism by which "growth forces an era split" (see
+ * `docs/contracts/CHANGELOG-conventions.md § Era splits`). The newest released
+ * section is exempt: a single release is allowed to be arbitrarily large and
+ * instead forces the split on the **next** bump, once it has become a prior
+ * entry. Without this exemption a single large catch-up release — e.g. a
+ * 376-commit bump whose `## [x.y.z]` section alone exceeds the cap — trips the
+ * gate with no era split that could ever bring it under, which is the exact
+ * failure this function fixes.
+ *
+ * Returns the same count as {@link current_era_body_size} when the era holds no
+ * released section yet (nothing to exempt).
+ */
+export function current_era_accumulated_body_size(lines: readonly string[] | null = null): number {
+    const all = lines ?? read_changelog_lines();
+    const spans = era_spans(all);
+    const current_idx = current_era_index(spans);
+    if (current_idx === null) {
+        return 0;
+    }
+    let next_era_line = all.length;
+    for (const span of spans) {
+        if (span.line_index > current_idx) {
+            next_era_line = span.line_index;
+            break;
+        }
+    }
+    const total = next_era_line - current_idx - 1;
+    if (total <= 0) {
+        return total;
+    }
+    // Topmost (newest) version heading in the era body.
+    let first = -1;
+    for (let i = current_idx + 1; i < next_era_line; i++) {
+        if (VERSION_HEADING_RE.test(all[i]!)) {
+            first = i;
+            break;
+        }
+    }
+    if (first === -1) {
+        return total; // no released section yet → nothing to exempt
+    }
+    // The next version heading marks where accumulated prior releases begin.
+    let next_ver = next_era_line;
+    for (let i = first + 1; i < next_era_line; i++) {
+        if (VERSION_HEADING_RE.test(all[i]!)) {
+            next_ver = i;
+            break;
+        }
+    }
+    const newest_section = next_ver - first;
+    return Math.max(0, total - newest_section);
+}
+
 /** Parse `M.N.x` into `[M, N]`; return null for archived labels. */
 export function parse_era_label(label: string): [number, number] | null {
     const m = ERA_LABEL_RE.exec(label.trim());

package/src/scripts/_lib/installed_lock.ts

@@ -429,23 +429,58 @@ function tuple_lt(a: [number, number, number], b: [number, number, number]): boo
   return false;
 }
 
-/** Read `version` from the package's own `package.json`. */
-export function current_package_version(repo_root?: string | null): string {
-  let root = repo_root ?? null;
-  if (root === null) {
-    // __file__ = src/scripts/_lib/installed_lock.ts → repo root is three
-    // levels up (_lib → scripts → src → root), i.e. parents[3] in Python.
-    const here = path.dirname(fileURLToPath(import.meta.url));
-    root = path.resolve(here, "..", "..", "..");
-  }
+/** Read a string `version` from `<dir>/package.json`, or null if absent/malformed. */
+function _read_package_version(dir: string): string | null {
   try {
-    const data = JSON.parse(fs.readFileSync(path.join(root, "package.json"), { encoding: "utf-8" }));
+    const data = JSON.parse(fs.readFileSync(path.join(dir, "package.json"), { encoding: "utf-8" }));
     const version = (data as { version?: unknown }).version;
     if (typeof version === "string" && version.trim()) {
       return version.trim();
     }
   } catch {
-    // Missing / unreadable / malformed package.json → fall through.
+    // Missing / unreadable / malformed package.json → null.
+  }
+  return null;
+}
+
+/**
+ * Walk upward from `start_dir` to the nearest `package.json` carrying a string
+ * `version`. Layout-independent: works in the source tree
+ * (`src/scripts/_lib/`), a per-file `dist/`, AND the esbuild single-file bundle
+ * (`dist/cli/agent-config.js`) where this module's runtime location differs
+ * from its source path. Returns null at the filesystem root.
+ */
+export function _find_package_version_upward(start_dir: string): string | null {
+  let dir = start_dir;
+  for (;;) {
+    const found = _read_package_version(dir);
+    if (found !== null) {
+      return found;
+    }
+    const parent = path.dirname(dir);
+    if (parent === dir) {
+      return null; // reached the filesystem root
+    }
+    dir = parent;
+  }
+}
+
+/**
+ * Read `version` from the package's own `package.json`.
+ *
+ * With no `repo_root`, the version is resolved by walking **upward** from this
+ * module's runtime directory to the nearest `package.json`. A previous fixed
+ * `parents[3]` hop assumed the source layout (`src/scripts/_lib/`); in the
+ * shipped esbuild bundle the module runs from `dist/cli/`, so three hops
+ * overshot the package root → `package.json` not found → `"0.0.0"` → a
+ * spurious "downgrade" refusal on `agent-config global` (lockfile recorded a
+ * real version, current package read as `0.0.0`). The upward walk is correct
+ * in both layouts.
+ */
+export function current_package_version(repo_root?: string | null): string {
+  if (repo_root != null) {
+    return _read_package_version(repo_root) ?? "0.0.0";
   }
-  return "0.0.0";
+  const here = path.dirname(fileURLToPath(import.meta.url));
+  return _find_package_version_upward(here) ?? "0.0.0";
 }

package/src/scripts/rdp_gate_classify.ts

@@ -0,0 +1,131 @@
+#!/usr/bin/env tsx
+/**
+ * RDP gate-time classification validation (road-to-rdp-frontier-polish Phase 1).
+ *
+ * The council's blocker on "keep-scoped" (2026-06-22): scoping the orchestrator
+ * to interdependent-multi-step work via agent self-assessment is only honest if
+ * the agent can classify a task AT GATE TIME — from the prompt alone, before
+ * doing the work. This script measures exactly that: the standard host classifies
+ * each labelled corpus prompt as multi-step vs single-turn, scored against the
+ * ground-truth `mechanism` label. High accuracy → the self-assessment gate is a
+ * real control, not vibes. Low accuracy → keep-scoped is not implementable.
+ *
+ * Direct Messages API via fetch (no SDK dep), reuses the key gate + cost
+ * estimate from skill_trigger_eval.ts. Dry-run by default; --confirm to spend
+ * (+ RDP_EVAL_ALLOW_NONTTY=1 on a non-tty caller that has confirmed cost).
+ *
+ * Usage: ./scripts-run src/scripts/rdp_gate_classify --corpus <file> --confirm
+ */
+import * as fs from 'node:fs';
+import { estimate_cost, load_anthropic_key } from './skill_trigger_eval.js';
+
+/** stdout helper — the lint config forbids `console.log` (allows warn/error). */
+const emit = (s: string): void => {
+    process.stdout.write(`${s}\n`);
+};
+
+const ANTHROPIC_URL = 'https://api.anthropic.com/v1/messages';
+const ANTHROPIC_VERSION = '2023-06-01';
+
+const GATE_SYSTEM = `You are an agent deciding, at the START of a task (before doing any work), whether
+to engage a heavyweight multi-step reasoning orchestrator. You see ONLY the task
+prompt. Classify the task into exactly one bucket using the rdp-gate criterion:
+
+- "multi-step": interdependent multi-step work where the ordering and handoffs
+  between steps matter (e.g. schema → API → job → UI; a migration with stages; a
+  cross-cutting refactor). The orchestrator helps here.
+- "single-turn": a single analytical or lookup question, a naming/explain/yes-no
+  ask, or a one-shot edit — work that does NOT decompose into ordered
+  interdependent steps. The orchestrator over-processes here.
+
+Respond with STRICT JSON only: {"bucket":"multi-step"|"single-turn","confidence":"high"|"medium"|"low"}`;
+
+interface Slot {
+    n: string;
+    slug: string;
+    mechanism?: string;
+    band: string;
+    prompt: string;
+}
+
+async function classify(apiKey: string, model: string, prompt: string): Promise<{ bucket: string; confidence: string } | null> {
+    const resp = await fetch(ANTHROPIC_URL, {
+        method: 'POST',
+        headers: {
+            'x-api-key': apiKey,
+            'anthropic-version': ANTHROPIC_VERSION,
+            'content-type': 'application/json',
+        },
+        body: JSON.stringify({
+            model,
+            max_tokens: 60,
+            system: GATE_SYSTEM,
+            messages: [{ role: 'user', content: `Task prompt:\n${prompt}` }],
+        }),
+    });
+    if (!resp.ok) throw new Error(`API ${resp.status}: ${(await resp.text()).slice(0, 300)}`);
+    const data = (await resp.json()) as {
+        content?: Array<{ type?: string; text?: string }>;
+        usage?: { input_tokens?: number; output_tokens?: number };
+    };
+    const text = (data.content ?? []).filter((b) => b.type === 'text').map((b) => b.text ?? '').join('');
+    const m = text.match(/\{[\s\S]*\}/);
+    if (!m) return null;
+    try {
+        return JSON.parse(m[0]) as { bucket: string; confidence: string };
+    } catch {
+        return null;
+    }
+}
+
+function expectedBucket(mechanism: string | undefined): string {
+    return mechanism === 'multi-stage' ? 'multi-step' : 'single-turn';
+}
+
+async function main(): Promise<number> {
+    const argv = process.argv.slice(2);
+    const args: Record<string, string | boolean> = {};
+    for (let i = 0; i < argv.length; i++) {
+        const t = argv[i];
+        if (t === undefined) continue;
+        if (t === '--confirm') args.confirm = true;
+        else if (t.startsWith('--')) {
+            const next = argv[i + 1];
+            if (next && !next.startsWith('--')) { args[t.slice(2)] = next; i++; } else args[t.slice(2)] = true;
+        }
+    }
+    const corpusPath = (args.corpus as string) ?? '';
+    const model = (args.model as string) ?? 'claude-haiku-4-5-20251001';
+    if (!corpusPath) throw new Error('--corpus <file> required');
+    const slots = (JSON.parse(fs.readFileSync(corpusPath, 'utf-8')) as { slots: Slot[] }).slots;
+
+    const est = slots.reduce((p) => p + estimate_cost(model, 250, 60), 0);
+    emit(`rdp-gate-classify · ${slots.length} prompts · ${model} · ~$${est.toFixed(4)}`);
+    if (!args.confirm) {
+        emit('DRY-RUN — no spend. Re-run with --confirm.');
+        return 0;
+    }
+    if (!process.stdin.isTTY && process.env.RDP_EVAL_ALLOW_NONTTY !== '1') {
+        throw new Error('Non-tty: set RDP_EVAL_ALLOW_NONTTY=1 only after confirming cost.');
+    }
+    const apiKey = load_anthropic_key();
+    let correct = 0;
+    const rows: Array<Record<string, unknown>> = [];
+    for (const s of slots) {
+        const r = await classify(apiKey, model, s.prompt);
+        const exp = expectedBucket(s.mechanism);
+        const got = r?.bucket ?? 'parse-error';
+        const ok = got === exp;
+        if (ok) correct++;
+        console.error(`  ${s.n} ${s.slug}: expected ${exp} · got ${got} (${r?.confidence ?? '-'}) ${ok ? 'OK' : 'MISS'}`);
+        rows.push({ slot: s.n, slug: s.slug, expected: exp, got, confidence: r?.confidence ?? null, ok });
+    }
+    const acc = Math.round((correct / slots.length) * 1000) / 10;
+    const out = (args.results as string) ?? '/tmp/rdp-gate-classify.json';
+    fs.writeFileSync(out, JSON.stringify({ model, n: slots.length, correct, accuracy_pct: acc, rows }, null, 2), 'utf-8');
+    emit(`\nGATE-TIME CLASSIFICATION ACCURACY: ${correct}/${slots.length} = ${acc}%  (model ${model})`);
+    emit(`Results: ${out}`);
+    return 0;
+}
+
+main().then((c) => process.exit(c), (e) => { console.error(e instanceof Error ? e.message : String(e)); process.exit(1); });