claudemd-cli 0.65.1 → 0.66.0

package/CHANGELOG.md

@@ -8,12 +8,32 @@ All notable changes to the `claudemd` plugin. This changelog tracks plugin artif
 - **Canonical spec version source**: `spec/CLAUDE.md` top-line title (`# AI-CODING-SPEC vX.Y.Z — Core`) + `spec/CLAUDE-changelog.md` top `##` entry.
 - **Plugin semver vs spec semver** are independent: plugin patch (0.2.0 → 0.2.1) may ship when spec is unchanged (this release); plugin minor (0.1.9 → 0.2.0) ships when spec minor updates (v0.2.0 shipped spec v6.10.0).
 
+## [0.66.0] - 2026-08-09
+
+Ships spec **v6.25.0** — a compression + relocation minor that restores attention-budget headroom. No hook, script, or command behavior change; the shipped artifact delta is the four spec files + manifests.
+
+- **spec: core 24793 → 23042 bytes (−7.1%), extended 49964 → 42999 bytes (−14%)** with zero rule-semantics change, zero section renames, and the enforcement partition unchanged (6 hook / 16 self / 2 both / 1 external). What left: version archaeology, rationale essays (rules keep one-line incident citations), duplicated enumerations, operator bookkeeping. What moved: §13.2 promotion gates, §13.3 gate tables, and the Recent-changes operator carry-forward now live in `OPERATOR.md` (which grows accordingly — it is not Agent-loaded, so the growth is free at runtime).
+- Basis: per-clause audit (`tasks/spec-lean-cut-candidates-2026-08-09.md`, local) against 30d rule-hits telemetry and incident memories; `hard-rules-audit.js` reported `demoteCandidates: []`, so every HARD rule survives verbatim at anchor level — `hard-rules.json` bumps `spec_version` only. An externally-drafted "lean v7" (core→11.4K, new caps, changelog restart) was evaluated and rejected; the decision record enumerates 8 rejected items with reasons.
+- Users upgrading: nothing to do. The spec reads ~1.1k tokens lighter per turn (core ≈6.0k → ≈4.9k estimated); every obligation, AUTH surface, and hook contract is unchanged. Revert path: pin the previous plugin version (0.65.2) or `CLAUDEMD_ALLOW_DOWNGRADE=1 node scripts/install.js` from a 0.65.2 checkout.
+
+## [0.65.2] - 2026-07-28
+
+The v0.65.0 red leg has an explanation, and the guidance v0.65.1 left behind was wrong. No shipped **code** changed: the tarball delta is the version field and this changelog entry (`CHANGELOG.md` is itself the largest shipped file, so "byte-identical apart from the version" — the first draft of this sentence — would have been self-falsifying). No hook, script, command or spec changed. It is released so the correction carries a version and reaches the marketplace channel, which serves the whole repository.
+
+- **fix: `contract.test.sh` splits `grep`'s exit-code spectrum.** The C-direction membership test collapsed exit 1 ("no match") and exit ≥2 ("failed to run" — fork / ENOMEM / signal) into a single else branch, so a transient runner spawn failure was reported as documentation drift. Exit ≥2 is now an explicit infrastructure fault. Same class as the repo's own recorded lesson that `grep` exit 2 is an error, not a zero-match — the gate had stepped on a rule this project already wrote down.
+- **Retraction: v0.65.1's "if the assertion flakes again, the awk extraction is the next suspect" was wrong — do not start there.** The load-bearing evidence is the re-run: attempt 1 (failure) and attempt 2 (success) of the identical v0.65.0 commit ran on the **same runner version and the same image** (`macos-26-arm64 / 20260720.0258.1`) and disagreed. Identical code, identical image, different outcome — that alone excludes every deterministic cause, and it also means v0.65.1's hardening was not what fixed it.
+- Two weaker arguments were used at first and are recorded here as *weaker*, not as co-equal proof: the failing leg parsed the same **47** documented pairs and ran the same **46** assertions as every passing leg (45 PASS + 1 FAIL, `110/111`), and the pair carries no regex metacharacter. The 47 is counted **before** the `sort -u`, so it cannot by itself exclude a collation drop; and the collision check was run on glibc, which is not the collator that was on trial. The re-run is what settles it.
+- v0.65.1's two hardenings (`LC_ALL=C` on the dedups, `grep -Fqx` for the lookup) stay: BSD `sort -u` deduplicating by collation and a literal compared as a BRE are real portability hazards on their own merits. They were simply not the cause, and the earlier entry now says so.
+- **What is proven, and what is not.** Proven: the gate could not distinguish "not found" from "did not run", and the failure is non-deterministic on a fixed image. Not proven: that a spawn failure actually occurred on that runner — GitHub does not log it, so the transient-fault account is inferred, not observed. Every mechanism *we enumerated* (parse, extraction, collation, pattern interpretation) is excluded; that is not the same as closure over an unenumerable space. The fix removes the conflation rather than asserting the fault.
+- The new branch is verified by a three-arm control rather than assumed reachable: a hit reports PASS, a miss reports drift, and a stub `grep` exiting 2 reports an infrastructure fault — three distinct outcomes.
+
 ## [0.65.1] - 2026-07-28
 
 Hotfix for a red `ci@main` leg on the v0.65.0 push. No hook, script or spec behavior changes — the failing assertion is in a test, and the released artifact is not affected.
 
 - **fix: `contract.test.sh` pins the two locale-sensitive operations in its (event, emitter) comparison.** BSD `sort -u` removes lines that *collate* equal rather than lines that are byte-equal, so under a UTF-8 locale it can silently drop a distinct pair; the membership test also used `grep -qx`, treating a literal pair as a basic regex. Both are now `LC_ALL=C`, and the lookup is `grep -Fqx`.
-- **Mechanism not established, and the comment says so.** v0.65.0's `ci@main` leg failed `C emitted 'fail-open' by 'ship-baseline' is NOT documented` on macos-latest/22 while **the same commit passed on the tag run** (all four legs), on both Linux legs, and on macos-latest/20. It did not reproduce locally across repeated runs or under `C` / `en_US.UTF-8` / `C.UTF-8`. A "the table row got too long" theory was checked and discarded — the `bypass-escape-hatch` row is 1691 chars against fail-open's 1102 and parses fine. So this hardens two real portability hazards rather than claiming a root cause. If the assertion flakes again, the extraction itself is the next suspect, not the collation.
+- **Correction (2026-07-28, after this version shipped): a defect that explains the failure was found, and this entry's closing guidance was wrong.** (The *mechanism* on the runner remains unproven — see v0.65.2.) Re-running the identical v0.65.0 commit — without the hardening below — passed, so the hardening is not what fixed it and the failure is not deterministic. The failing leg parsed the same **47** documented pairs and ran the same **46** assertions as every passing leg; no collation collision exists among those 47 under `C` or `en_US.UTF-8`; the pair contains no regex metacharacter. That exonerates the awk extraction this entry named as "the next suspect" — **do not start there.** What remains is that the C-direction membership test collapsed `grep` exit 1 ("no match") and exit ≥2 ("failed to run") into one branch, so a transient runner spawn failure was reported as documentation drift. Same class as the repo's own recorded lesson that grep exit 2 is an error, not a zero-match. The exit-code spectrum is now split. The two hardenings below remain — they are real portability hazards on their own merits — but they were not the cause.
+- **Mechanism not established at the time, and the comment said so.** v0.65.0's `ci@main` leg failed `C emitted 'fail-open' by 'ship-baseline' is NOT documented` on macos-latest/22 while **the same commit passed on the tag run** (all four legs), on both Linux legs, and on macos-latest/20. It did not reproduce locally across repeated runs or under `C` / `en_US.UTF-8` / `C.UTF-8`. A "the table row got too long" theory was checked and discarded — the `bypass-escape-hatch` row is 1691 chars against fail-open's 1102 and parses fine. So this hardens two real portability hazards rather than claiming a root cause. If the assertion flakes again, the extraction itself is the next suspect, not the collation.
 - Worth stating plainly: v0.65.0's tag run passed all four matrix legs and `npm-publish` succeeded, so **0.65.0 on npm is not broken** — no deprecate. The red run is the duplicate `ci@main` run that the atomic ship flow fires on the same commit.
 
 ## [0.65.0] - 2026-07-28

package/README.md

@@ -1,12 +1,12 @@
 # claudemd
 
-> A **personal AI-coding discipline harness**: one developer's opinionated **AI-CODING-SPEC v6.24**, encoded as Claude Code shell hooks and shipped with the plugin. Built and dogfooded on my own repos — fork and adapt, don't adopt wholesale.
+> A **personal AI-coding discipline harness**: one developer's opinionated **AI-CODING-SPEC v6.25**, encoded as Claude Code shell hooks and shipped with the plugin. Built and dogfooded on my own repos — fork and adapt, don't adopt wholesale.
 
 [![CI](https://github.com/sdsrss/claudemd/actions/workflows/ci.yml/badge.svg?branch=main)](https://github.com/sdsrss/claudemd/actions/workflows/ci.yml)
 [![npm](https://img.shields.io/npm/v/claudemd-cli.svg)](https://www.npmjs.com/package/claudemd-cli)
 [![License: MIT](https://img.shields.io/badge/license-MIT-blue.svg)](LICENSE)
 
-claudemd plugs into the Claude Code hook system to **block commits, pushes, and bash commands** that violate AI-CODING-SPEC v6.24 — banned vocabulary in commit messages, `rm -rf $VAR` without variable validation, ship-on-red-CI, unread `MEMORY.md` entries during release flows, and more. The spec itself (`CLAUDE.md` + `CLAUDE-extended.md` + `CLAUDE-changelog.md` + `OPERATOR.md`) ships with the plugin and installs into `~/.claude/`, so the rules Claude Code reads at session start match the rules the hooks enforce. (`OPERATOR.md` is the human-only spec-maintenance handbook — Agent-loaded files are the CLAUDE trio.)
+claudemd plugs into the Claude Code hook system to **block commits, pushes, and bash commands** that violate AI-CODING-SPEC v6.25 — banned vocabulary in commit messages, `rm -rf $VAR` without variable validation, ship-on-red-CI, unread `MEMORY.md` entries during release flows, and more. The spec itself (`CLAUDE.md` + `CLAUDE-extended.md` + `CLAUDE-changelog.md` + `OPERATOR.md`) ships with the plugin and installs into `~/.claude/`, so the rules Claude Code reads at session start match the rules the hooks enforce. (`OPERATOR.md` is the human-only spec-maintenance handbook — Agent-loaded files are the CLAUDE trio.)
 
 A standalone CLI (`npx claudemd-cli`) reuses the same `banned-vocab.patterns` source for git pre-commit hooks, GitHub Actions, and other agents that don't run inside Claude Code.
 
@@ -60,7 +60,7 @@ Verify in one command (Linux): `node --version && jq --version && gh --version &
 | 15 shell hooks | `banned-vocab-check` · `pre-bash-safety-check` · `ship-baseline-check` · `residue-audit` · `memory-read-check` · `memory-prompt-hint` · `sandbox-disposal-check` · `session-start-check` · `session-extended-read` · `session-summary` · `session-end-check` · `transcript-vocab-scan` · `transcript-structure-scan` · `version-sync` · `mem-audit` |
 | 16 slash commands | `/claudemd-install` · `/claudemd-status` · `/claudemd-update` · `/claudemd-refresh` · `/claudemd-audit` · `/claudemd-toggle` · `/claudemd-doctor` · `/claudemd-analyze` · `/claudemd-uninstall` · `/claudemd-rules` · `/claudemd-clean-residue` · `/claudemd-sparkline` · `/claudemd-sampling-audit` · `/claudemd-bypass-audit` · `/claudemd-design-adopt` · `/claudemd-statusline` |
 | 1 standalone CLI | `claudemd-cli lint` · `claudemd-cli audit` ([npm: `claudemd-cli`](https://www.npmjs.com/package/claudemd-cli)) |
-| Spec v6.24 | `~/.claude/CLAUDE.md` · `CLAUDE-extended.md` · `CLAUDE-changelog.md` · `OPERATOR.md` (backup-before-overwrite) |
+| Spec v6.25 | `~/.claude/CLAUDE.md` · `CLAUDE-extended.md` · `CLAUDE-changelog.md` · `OPERATOR.md` (backup-before-overwrite) |
 | StatusLine (opt-out) | PS1-style line — `user@host:dir (branch) Model [ctx:N% · 5h:N% · 7d:N%]` (`dir` = cwd basename; context / 5-hour quota / weekly quota, all **used %**, read from Claude Code's `rate_limits` payload; quota segments auto-hide when the data is absent, or force-hide with `DISABLE_STATUSLINE_QUOTA=1`) — wired into `~/.claude/settings.json` on install **only when the slot is empty**; an existing statusline is left untouched. Skip entirely with `CLAUDEMD_NO_STATUSLINE=1`. Manage via `/claudemd-statusline`. |
 
 Install backs up a hand-written `~/.claude/CLAUDE.md` (any file without the `# AI-CODING-SPEC` H1) to `~/.claude/backup-<ISO>/` before overwriting (last 5 kept automatically). An already-installed claudemd spec is overwritten **without** a backup — deliberate (v0.23.11): the sole backup is always your own content, so `restore` can never return a stale spec instead. Uninstall offers `keep / restore / delete`; `delete` requires an extra confirmation.
@@ -352,7 +352,7 @@ claudemd/
 ├── commands/                 # 16 slash-command markdown files
 ├── bin/                      # standalone CLI entrypoint (claudemd-lint.js → `npx claudemd-cli` on npmjs.org)
 ├── scripts/                  # 18 Node.js scripts + scripts/lib/ (single-source registry, lint, etc.)
-├── spec/                     # shipped v6.24.0 CLAUDE*.md trio + OPERATOR.md + hard-rules.json manifest
+├── spec/                     # shipped v6.25.0 CLAUDE*.md trio + OPERATOR.md + hard-rules.json manifest
 ├── tests/                    # hook shell tests + Node.js tests + integration + fixtures
 ├── docs/                     # ADDING-NEW-HOOK.md + RULE-HITS-SCHEMA.md + superpowers/
 └── .github/workflows/        # ci.yml (ubuntu+macOS × node 20) + npm-publish.yml (tag-triggered)

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "claudemd-cli",
-  "version": "0.65.1",
+  "version": "0.66.0",
   "description": "Standalone CLI for §10-V banned-vocab + transcript scanning. Companion to the claudemd Claude Code plugin (github.com/sdsrss/claudemd) for use in git pre-commit hooks, GitHub Actions, and other agents.",
   "type": "module",
   "bin": {