@chrono-meta/fh-gate 2.6.0 → 2.7.0

package/.claude/rules/fh_4axis_gate.md

@@ -190,9 +190,32 @@ controls: n/a — no measurement in this delta (<reason>)
 ```
 
 **`standpoint:`** — closed enum, canonical spec in
-`knowledge/shared/harness-core/field_verdict_crossfamily_gate.md §7`. **Still validated by nothing**
-— zero hook lines, no fixture suite. Recorded here so the gap is visible from the rules side rather
-than only from the hook's absence.
+`knowledge/shared/harness-core/field_verdict_crossfamily_gate.md §7`. 🟥 **CORRECTED 2026-08-20, then CORRECTED AGAIN the same hour.** This used to read "Still validated
+by nothing — zero hook lines, no fixture suite", which is FALSE: `validate_standpoint_leg()` lives in
+`templates/.git-hooks/pre-commit` (grep the function name — **line numbers are deliberately not cited
+here; the first version of this fix cited `:798`/`:1575` and a commit landed the same hour that moved
+them to `:878`/`:1665`**), and `scripts/test_marker_standpoint_lanes.sh` is wired through
+`scripts/selfcheck.sh`.
+🟥 **But the first correction over-shot, and cross-family caught that too.** It said "the value enum
+IS closed and the grounds ARE required". Only the first half holds. Measured by varying ONE variable
+at a time — the original known-pair varied two and mis-attributed the result:
+
+```
+standpoint: tier2(qasp) — ran <cmd>, saw <out>   rc=0   ✅
+standpoint: tier2(qasp)          (no grounds)    rc=0   ⚠️  warns, does NOT block
+standpoint: tier2                (no parens)     rc=1   ❌  not an enum member
+standpoint: banana(qasp)                         rc=1   ❌  not an enum member
+```
+
+So: **the enum is closed and enforced; the execution grounds for `tier2`+ are ADVISORY.** A marker can
+still record `tier2` without naming a command and pass with a warning — that is the real remaining
+gap, and it is narrower than "validated by nothing" and wider than "grounds are required". Neither
+earlier sentence was accurate, and the accurate one required varying one variable at a time.
+🟥 **This same false claim stood in FIVE places**, not three — `CLAUDE.md` (twice), here, `AGENTS.md`,
+and `field_verdict_crossfamily_gate.md`. Fixing one and stopping is the half-fix propagation failure;
+fixing three and stopping was the same failure one round later. The question is never "is this
+sentence wrong" but **"where else does this sentence live"** — and the answer came from a
+different-family reviewer, not from me.
 
 **`thirdparty:` — ⓓ3자 대면의 자기 필드 (2026-08-17 신설).** ⓑ가 `standpoint:` 를 갖는 것과
 같은 형태다: `axes-run` 에는 포인터(`ⓓ=→thirdparty`)만 두고 값은 이 필드가 나른다.

package/.claude-plugin/marketplace.json

@@ -11,13 +11,13 @@
   "plugins": [
     {
       "name": "fh-meta",
-      "version": "2.6.0",
+      "version": "2.7.0",
       "description": "New in 2.2.0: BREAKING (gate): chamber step 6 now reads ACTUAL.md, not BUDGET.md — an in-flight chamber run whose actual cost sits in BUDGET.md blocks until the ACTUAL: line moves to tracks/_chamber/<slug>/ACTUAL.md (the runner prints the path). Why: BUDGET.md's pre-verdict hash IS the ordering witness, and step 6 hard-blocked until that same file changed, so every run that reached COMPLETE necessarily mutated a witnessed artifact and verify returned TAMPERED — the chamber's promotion condition was unsatisfiable by construction, not by strictness. Two roles (immutable witness / post-verdict calibration sink) had collided in one file; each was correct alone, so neither side's code showed the conflict. Also: ko-tech-writer Step 2/4-b scans are now calibration-backed (known-pair fixtures + reproducible command, shipped) — discrimination is proven, 'zero residue' is explicitly NOT; chamber lane suite 12 -> 33 including the runner x witness seam no test covered; chamber_run.sh now teaches the two-commit discipline (gate hashes and verdict hash must land in separate commits/PRs — it previously advised the opposite). New in 2.1.0: BREAKING (gate): `crossfamily: declined` in an Axes 2-3 marker now requires grounds naming a record path that RESOLVES on disk — bare `declined`, and `declined` justified by author judgment, are blocked at commit. Remedy: cite where the operator decision lives (e.g. `.. — operator declined sidecars, per knowledge/shared/rules/operational_adaptation.md`), or use `DEGRADED_PANEL_UNUSED` if a panel was reachable and you chose not to recruit it — which is what author judgment actually is. `declined` was the only enum value with no grounds requirement; a cross-family review then broke the first (vocabulary-grep) fix three ways — self-validating on the value's own token, vacuous keyword passes, and over-blocking real declinations in natural prose — so the check asserts a resolvable record instead of words. Also: standpoint axis gains `tier1b` (a STATIC read of a target repo, executed nothing) plus a decide-in-order procedure, after blind floor-tier sims graded pure cold-reads as `tier2` three rounds running; steel-quench Wave 1's sixth angle (gate-locality) gains the output-template row it never had, so a mandatory angle stops being structurally unreportable; verify-bidirectional gains category 5 (prescriptive doctrine statement); Sister Asset Protocol gains an active-adoption trigger; new resident doctrine — Mechanization Boundary, Local Execution First, Skeleton-not-Muscle, Expedition track, and this package's versioning policy. Hub meta-operations toolkit — 35 skills + 7 agents. New in 2.0.1: harness-doctor cadence hook, portability lint wired into pre-commit, branch_claim.sh claim-count-vs-tree-count warning, louder confidentiality-scan fail-open notice, fh-gate.sh missing-package.json survival, identity ① reclassified 🟢 (cross-harness adapters + relay argument channel). New in 1.4.53: `fh-codex-doctor` (npm bin) — Codex adapter drift scanner; reads the documented M1/M2/M3 skill tier map + skill/agent source and reports codex-native/adapter-required/claude-native/unclassified per unit, wired into `npm test`/`prepublishOnly` (fail-closed on unclassified Claude-native primitives). New in 1.4.49: steel-quench gains Step 0.6 Verdict-Invariance Probe (groundedness axis — a load-bearing judged gate's verdict must track behavior, not rubric phrasing; measured flip-count over cross-family paraphrases; arXiv:2605.06161 Policy Invariance anchor); multi_model_sidecar_strategy §Vendor-native harness (a model is strongest in its own vendor CLI — Claude/CC, GPT/codex, Gemini/Antigravity; a universal router degrades all of them, so it stays an autocomplete/QA sidecar, never orchestration); predelete_check.sh fail-closed rewrite; memory-hygiene A-TMA anchor. New in 1.4.48: phantom-quench + steel-quench gain external frontier anchors (arXiv:2607.02052 package-hallucination; arXiv:2607.02057 prompt-coverage-adequacy); README model-flat claim reframed from a per-release point-curve to structural invariants (operation flattens across tiers; depth tier-order fixed within a generation). New in 1.4.47: onboarding step ① surfaces the Mode D companion-store session-start load in the auto-read salience anchor (previously only in the local binding + rules, so a greeting could skip the load). New in 1.4.46: context-doctor command-output axis (route to rtk/proxy for verbose CLI stdout, complementing .claudeignore; risk-gated to token-scarce envs). New in 1.4.41: context-doctor 2026 trigger vocab (context engineering/rot/collapse) + phantom-citation hardening; hub measurement-integrity-checklist (cross-model measurement pre-flight: display-name pin/reps≥3/discriminating probe). New in 1.4.40: install-wizard queryable-wiki scaffold (INDEX + session-start read + R/W/C ingest). New in 1.4.39: auto-decorrelation (cross-family verifier sidecar recruitment) + video-ingest (capability-routed video ingestion). New in 1.4.x: verify-axis check-class taxonomy (mandatory-pass/measured/judged), no-reinvention Tier-0 inventory, 7-class failure taxonomy, Destructive-Op Gate, Wave-T (Temper), tier-floor governance, Mode D Model Notice, FC consent lane, default-Sonnet guidance. New in 1.3.0: public-surface-audit, field-harvest Mode B auto-trigger, 4-axis gate scope ext. Validated cross-CLI: Claude Code, Codex, Gemini.",
       "source": "./plugins/fh-meta"
     },
     {
       "name": "fh-commons",
-      "version": "2.6.0",
+      "version": "2.7.0",
       "description": "Project-agnostic utility skills — 5 skills (convergence-loop · deliberation · mcp-circuit-breaker · token-budget-gate · ko-tech-writer) + 1 agent (quench-challenger). Domain-independent utilities transplantable into any project.",
       "source": "./plugins/fh-commons"
     }

package/AGENTS.md

@@ -127,9 +127,14 @@ Because non-Claude runtimes do not auto-load Claude path rules, apply these rule
    2 of the 4 circled-key markers on disk are dated 2026-08-10 and carry the OLD meanings. Aligning
    the notation still helps going forward; it does not work backwards.
    `standpoint:` remains the canonical field for ⓑ (the `axes-run` entry
-   is only a pointer to it, and a pointer at an empty field is blocked); **its value enum is still
-   validated by nothing** — that is the one remaining gap, and it is not the same thing as the axis
-   being unmechanized. Format spec: `.claude/rules/fh_4axis_gate.md §Marker axis fields`.
+   is only a pointer to it, and a pointer at an empty field is blocked). 🟥 **CORRECTED 2026-08-20.** This used to say
+   "its value enum is still validated by nothing" — FALSE. `validate_standpoint_leg()` in the
+   pre-commit hook enforces a closed enum. Measured by varying one variable at a time:
+   `standpoint: banana(qasp)` → **blocked (enum)** · `standpoint: tier2` without parens → **blocked
+   (enum)** · `standpoint: tier2(qasp)` with no execution grounds → **passes with a warning** ·
+   with grounds → **passes**. So the enum is enforced and the `tier2`+ execution grounds are
+   **advisory** — a first version of this correction said grounds were required, which over-shot.
+   The hook also cannot check whether `tier2` is **true**. Those two, together, are the gap. Format spec: `.claude/rules/fh_4axis_gate.md §Marker axis fields`.
    (Two drift corrections landed here on 2026-08-17: first this sentence said "four" while its own
    next clause described the +1 — caught by the session-close ④-b CC↔Codex parity check — and then
    the machine layer moved to six the same day.)
@@ -144,6 +149,22 @@ Because non-Claude runtimes do not auto-load Claude path rules, apply these rule
    `❌`. Execution is the half with no substitute; see `field_verdict_crossfamily_gate.md §7`. Read it before recording; §1-c
    holds the sample limits — read it before citing. This record is self-attested and has no hook
    behind it; it is closed by a different-family reader, not by writing it more carefully.
+   🟥 **`axis2-defense:` — a marker field added 2026-08-20 that a Codex-side author WILL hit.**
+   It is required when the marker records `floor-status: sonnet-floor` or `below-floor`, and it carries
+   three sub-answers (on one line, or across continuation lines — the hook reads both) about **your own findings and numbers**:
+   `axis2-defense: reproducibility=<exact command or file:line another session runs> fairness=<reps,
+   inputs and environment named for BOTH arms> estimation-layer=<per number: measured | estimate |
+   quotation; for a measurement, what showed the instrument works on this target>`.
+   The hook (`validate_defense_leg`) checks **presence · completeness · non-vacuity** — `ok`/`yes`/
+   `n/a` is rejected as a filled form rather than an answer, and all three sub-answers must exist.
+   It cannot check whether the answers are true. Why it exists: measured n=6 in the origin field, a
+   floor-tier pass runs the attack angles without defect but does **not** spontaneously ask these
+   three; that is a checklist gap, not a capability gap, and a harness whose behaviour depends on
+   which model drives it is defective by the Sonnet-Floor doctrine.
+   Fixtures: `scripts/test_marker_defense_lanes.sh`. Spec: `plugins/fh-meta/skills/steel-quench/SKILL.md`
+   §Wave 1-D. **Recorded here because a rule living in only one entry point is invisible to the other
+   runtime** — that is the gate-locality principle, and this line is it being applied rather than cited.
+
 8. **Branch-surface claims:** GitHub branch protection is two independent layers — legacy
    protection and rulesets coexist, and the strictest wins. Read both
    `/repos/{owner}/{repo}/branches/{branch}/protection` and

package/CLAUDE.md

@@ -376,8 +376,18 @@ claim 이 `main` 인 동안 실제 HEAD 는 peer 브랜치였다 — 두 세션
 **남은 잔여 둘, 이름으로 남긴다**: ⓐ 게이트는 **커밋 시점**이라 `switch -c` 사고 자체는 못 막는다
 (그래서 위 두 줄이 여전히 사람 몫이다 — [[feedback_gate_binding_point_not_check_point]])
 ⓑ 반대 방향 — 내가 브랜치를 쥔 동안 남이 되돌리고 커밋하면 **내 staged 파일이 index 에 살아
-있다.** 실측 사례에서 안 섞인 것은 그 세션이 파일 지정 `add` 를 썼기 때문이지 게이트 덕이 아니다.
-⇒ **공유 체크아웃에서 `git add -A` 금지**([[feedback_shared_checkout_ops_touch_others_work]]).
+있다.** 🟥 **2026-08-21 실제로 났고, 섞은 세션은 파일 지정 `add` 를 썼다**(초판은 여기 「파일
+지정 add 를 썼기 때문에 안 섞였다」고 적었는데 거짓이다). 파일 지정 add 는 «내가 무엇을
+**추가**하나»만 통제하고 **index 에 이미 있는 것은 못 막는다.** ⇒ `git add -A` 금지는 필요조건
+이지 충분조건이 아니다. **커밋을 한 호출로 묶고 그 안에서 둘을 확인한다** — 확인은 점이고 위험은
+구간이라, 사이에 턴이 끼면 창이 다시 생긴다(같은 날 두 세션이 시점을 각각 `switch` 직전/직후로
+달리 골랐는데 **둘 다 뚫렸다**):
+```
+B=$(git branch --show-current); [ "$B" = "<내 브랜치>" ] || exit 1
+git diff --cached --name-only    # 내 것만 있나 — 남의 것은 restore --staged
+git commit …
+```
+([[feedback_shared_checkout_ops_touch_others_work]] · [[feedback_gate_binding_point_not_check_point]])
 
 **Integration branch is PR-only** (operator decision 2026-07-20). Never `git push origin main`
 directly. Normal path: `git switch -c <branch>` → push the branch → `gh pr create` → after review
@@ -625,17 +635,24 @@ English with FH's own persona/viewpoint sense of "standpoint" (`fh-meta:beginner
 `expert`) — a different axis (which persona reviews, not whose repo is ground truth); kept as-is,
 not renamed, but do not conflate the two. 🟥 **CORRECTED 2026-08-20 — this paragraph used to say
 `standpoint:` was "Prose-only today — no pre-commit hook or fixture suite validates this field yet".
-That is FALSE and was false in this same file**: `validate_standpoint_leg()` is defined at
-`templates/.git-hooks/pre-commit:798` and called at `:1575`, and its fixture suite
-`scripts/test_marker_standpoint_lanes.sh` is wired through `scripts/selfcheck.sh:531`. §자기 대조
+That is FALSE and was false in this same file**: `validate_standpoint_leg()` lives in
+`templates/.git-hooks/pre-commit` and is called from the marker block, and its fixture suite
+`scripts/test_marker_standpoint_lanes.sh` is wired through `scripts/selfcheck.sh`. (🟥 **Grep the
+names, do not trust line numbers** — the first version of this correction cited `:798`/`:1575` and a
+commit landed the same hour that moved them to `:878`/`:1665`. A hardcoded anchor in prose is a
+phantom waiting for the next edit.) §자기 대조
 above already said so (PR #429), so **one file carried both claims at once** and a reader landed on
 whichever they reached first. Found by the residency-ledger pass, not by a lane — no check compares
 a rule's self-description against the machinery it describes, which is why a stale "we have not
 built this yet" is the quietest form of drift: it reads as honest modesty and it suppresses use of a
-control that already exists. **What is validated is FORM, not truth** — the hook checks that the
-value is inside the closed enum and that its grounds are non-empty; whether `tier2` is *true* is
-still self-attested, exactly as with `crossfamily:`. That residual is real and unchanged; it is the
-sentence above that was wrong, not the caution. Three artifacts, one carrying two
+control that already exists. **What is validated is the ENUM** — measured by varying ONE variable at a
+time, because the first version of this correction varied two and mis-attributed the result:
+`banana(qasp)` → blocked (enum) · `tier2` without parens → blocked (enum) · `tier2(qasp)` with **no**
+execution grounds → **passes with a warning** · with grounds → passes. 🟥 So the first fix's claim
+that "grounds are non-empty" are checked **over-shot, and a different-family reviewer caught it**:
+the `tier2`+ execution grounds are **advisory**. Two residuals remain and both are real — grounds are
+not enforced, and whether `tier2` is *true* is still self-attested. What was wrong was only the claim
+that nothing validated the field at all. Three artifacts, one carrying two
 independent trials (forge-harness PR #368, a sibling field harness's PR #8 reps=3 and its
 known-answer trial, qasp-dev PR #161 as adjacent corroboration) crossed this repo's own evidence
 bar the same day this was formalized — including one caught by this session's own qasp PR #161
@@ -824,7 +841,8 @@ deleted branch was carrying).
 **When this gate fires** — *before* any of: branch deletion (local or remote), history rewrite /
 force-push, scrub of tracked history, bulk deletion of session records / tracks content.
 
-1. **Enumerate (measured)**: `bash templates/predelete_check.sh <repo> [base]` — per branch: commits
+1. **Enumerate (measured)** — 🟥 **this step is run BY A HUMAN; no hook executes it** (measured
+   2026-08-20, residency-ledger pass): `bash templates/predelete_check.sh <repo> [base]` — per branch: commits
    off base + unique paths. Verdicts: SAFE (fully merged) · CHECK (0 unique paths but commits off
    base — shared files may hold *newer* content, e.g. an unmerged session card) · REVIEW (unique
    paths — recovery mandatory).
@@ -845,6 +863,15 @@ prose gate is now stopped); it does **not** close the injected/adversarial one
 or errors, this irreversible surface **fails closed** — the pre-push hook blocks (enumerate by hand or
 take the explicit `DESTRUCTIVE_OP_OK=1` override); a tooling-down enumerate step never silently degrades
 into "just delete it."
+🟥 **Read that precisely — the floor is the HOOK, not this script.** `templates/.git-hooks/pre-push`
+implements the per-ref verdict **inline**; it does not call `predelete_check.sh`. Measured 2026-08-20
+(control: the same scan finds `session_close_check` wired in that hook): every in-repo reference to
+`predelete_check.sh` is a **mention, not an execution** — `pre-push:408` lists the path inside a *grep
+pattern*, `destructive_pre_gate.sh:191` *prints the command* as advisory text, and `selfcheck.sh:192`
+runs `bash -n` on it. So the script is the **operator's enumerate tool**, and step 1 above is a human
+step that the machinery reminds you of rather than performs. Stating it the other way round is the
+"prose-invoked floor" the 4-axis marker spec calls M-tier when a rule claims a floor its script has
+no caller for — this section does not make that claim, and this note keeps it from drifting into one.
 
 > **Detail**: See `knowledge/shared/harness-core/claude_md_gate_details.md §Destructive-Op-Hook-Coverage`
 > — the per-ref verdict mechanics, what the hook does/does not close (honest scope + adversarial residual),
@@ -1005,7 +1032,7 @@ Self-healing is not only FH-self-dev (Mode D 4-axis) and reactive (`verify-bidir
 
 ## Agent Dispatch Operation (FH cwd-Based)
 
-> **Runtime authority (canonical):** one explicit governor per context + capability-routed sidecars; sidecar findings are evidence candidates, not terminal verdicts, until source-closed by the governor *via a mechanical anchor* — never governor agreement alone. CC=action/governor · Codex=repo-grounded audit sidecar · Gemini/agy=breadth/multimodal sidecar · other runtimes=portable `AGENTS.md` entrypoint only. Full doctrine + Maintenance-Cost Rule: `knowledge/shared/harness-core/multi_model_sidecar_strategy.md §Runtime Authority`.
+> **Runtime authority (canonical):** one explicit governor per context + capability-routed sidecars; sidecar findings are evidence candidates, not terminal verdicts, until source-closed by the governor *via a mechanical anchor* — never governor agreement alone. 🟥 **A sidecar audits; it does not WRITE to the target tree** — findings and at-most a proposed patch as text, applied by the governor (measured 2026-08-21: an auditor sidecar edited the tree and its fix introduced a self-referential fail-open that 41 lanes passed). CC=action/governor · Codex=repo-grounded audit sidecar · Gemini/agy=breadth/multimodal sidecar · other runtimes=portable `AGENTS.md` entrypoint only. Full doctrine + Maintenance-Cost Rule: `knowledge/shared/harness-core/multi_model_sidecar_strategy.md §Runtime Authority`.
 
 **Isolated delegation is a component of the identity, not an optional extra** (operator decision,
 2026-08-08). FH/PMH are defined as governor + orchestrator; a harness that cannot dispatch is a
@@ -1317,7 +1344,7 @@ Closing phrase detected ("wrap up", "done", "good work", "end session", etc.)
 
        | Bump | Reserved for (operator's own wording, 2026-08-17) |
        |---|---|
-       | **major** `+1.0.0` | **any one of three**: ⓐ **완전히 새로 지음** — rebuilt from scratch, not extended · ⓑ **정체성이 확립됨** — an identity of the five (+Ⓑ) actually standing 🟢, not progressing toward it · ⓒ **기능이 혁신적으로 변경되거나 늘어남** — a capability *class* appears or is replaced, not a capability instance. 🟥 **Never** for tightening a gate that already existed |
+       | **major** `+1.0.0` | **any one of three**: ⓐ **완전히 새로 지음** — rebuilt from scratch, not extended · ⓑ **정체성이 확립됨** — 🟥 **다섯이 «전부» 🟢** 인 순간이지 하나가 🟢 로 올라선 순간이 아니다(운영자 결정 2026-08-21). 초판은 *"an identity of the five … actually standing 🟢"* 였고 **「하나만 초록이어도 major」로 읽혔다** — 실제로 그날 ②가 🟢 로 판정되면서 3.0.0 후보로 올라왔고, 그 애매함이 그때 닫혔다. 🟥 그리고 **정체성 등급은 npm 이 나르는 신호가 아니다** — 그건 `identity-v*` 계보의 사건이고, npm 이 또 나르면 같은 날 고친 「두 계보 한 이름」 결함을 번호에서 재생산한다. ⇒ major-ⓑ 는 **`identity-v1.0.0` 과 같은 사건**을 가리킨다 · ⓒ **기능이 혁신적으로 변경되거나 늘어남** — a capability *class* appears or is replaced, not a capability instance. 🟥 **Never** for tightening a gate that already existed |
        | **minor** `+0.1.0` | 미들급 — new assets, new gate lanes, doctrine that changes behavior; **including changes that break a consumer's gate acceptance**, which then carry a mandatory `BREAKING (gate):` line |
        | **patch** `+0.0.1` | 트리비아급 — fixes, wiring, docs that change no behavior |
 

package/README.ja.md

@@ -18,24 +18,48 @@
 </p>
 
 <p align="center">
-  <sub>役に立ったら ⭐ が他の人の発見につながります。</sub>
+  <b>やりたいことを頼んでください。同じ依頼が繰り返されたら、それ自体を作りましょうと先に提案します。</b>
 </p>
 
 <p align="center">
-  <b>あなたの Claude Code プロジェクトを鍛えて — 通せば、より速く仕上がります。</b><br>
-  実務者の<b>メタハーネス (meta-harness)</b> — あなたのプロジェクトハーネスたちが暮らす銀河。<br>各プロジェクトの<b>床 (floor)</b> を上げ（設定をハーネス化）、<b>天井 (ceiling)</b> を上げた上で（作業を加速）、その利得をポートフォリオ全体に複利で積み上げます。
+  すでに Claude Code へ同じことを繰り返し伝えているはずです。走らせる検査、守るべきルール、
+  変更が満たすべき形。<br>
+  <b>forge-harness はそれを再利用できるものに変えます。</b> リポジトリの中に住み、自分から発火する
+  スキル・ゲート・エージェントとして。<br>
+  スキルはあえて汎用のままにし、使ううちにあなたの事例へ合わせて鍛えます。
+  <b>同じ形が何度も戻ってきたら、そこで出荷を提案します</b> — 独立したスキルとして、または独立したハーネスとして。
 </p>
 
-<p align="center">
-  <b>品質が梃子であり、速度はその結果です。</b> あらゆる変更はゲートを通って自らの値打ちを証明します —<br>敵対的 (adversarial) · ファントム (phantom) · 回帰 (regression) — そして<i>それ</i>が次の変更をより速くします。
-</p>
+---
+
+## 2分で試せます — この文書を読み切る必要はありません
+
+```bash
+claude plugin marketplace add https://github.com/chrono-meta/forge-harness.git
+claude plugin install -s user fh-meta@forge-harness
+git clone https://github.com/chrono-meta/forge-harness.git ~/projects/forge-harness
+cd ~/projects/forge-harness && claude
+```
+
+**そして `hi` と入力してください。** 番号付きのメニューが出て、そこからはツールが案内します。
+入口を選んでいくつか答えれば、インストールウィザードまで代わりに実行します。この線から下は
+必要になったときに見る参考資料であり、始める前の宿題ではありません。
+
+**増幅するもの**: 試行の回数です。試行錯誤があなたから離れ、並列で回ります。
+**増幅しないもの**: モデルの天井です。ハーネスはモデルを自分の天井まで引き上げるだけです。
+**確かめ方**: 自分の評価を公開します。五つのアイデンティティを正直に、
+[リリース](https://github.com/chrono-meta/forge-harness/releases)ごとに記します。緑でない欄は、
+まだ足りていない実際の実行が何かを名前で述べます。
+
+---
 
 <p align="center">
-  <i>フォークしてください。名前を変えてください。あなたのものにしてください。</i>
+  <img src="docs/pillars.svg" alt="FORK - ADAPT - COLLABORATE - EMPOWER" width="680">
 </p>
 
 <p align="center">
-  <img src="docs/pillars.svg" alt="FORK · ADAPT · COLLABORATE · EMPOWER" width="680">
+  <b>品質が梃子であり、速度はその結果です。</b> <i>フォークしてください。名前を変えてください。あなたのものにしてください。</i><br>
+  <sub>役に立ったら ⭐ が他の人の発見につながります。</sub>
 </p>
 
 <p align="center">

package/README.ko.md

@@ -18,24 +18,48 @@
 </p>
 
 <p align="center">
-  <sub>도움이 됐다면 ⭐ 하나가 다른 사람이 찾는 데 도움이 됩니다.</sub>
+  <b>필요한 걸 시키세요. 같은 요청이 반복되면, 그걸 만들어 드리겠다고 먼저 제안합니다.</b>
 </p>
 
 <p align="center">
-  <b>당신의 Claude Code 프로젝트를 벼려서 — 통과시키면, 더 빠르게 나옵니다.</b><br>
-  실무자의 <b>메타하네스</b> — 당신의 프로젝트 하네스들이 사는 은하계.<br>각 프로젝트의 <b>바닥(floor)</b>을 올리고(설정을 하네스화) <b>천장(ceiling)</b>을 올린 뒤(작업을 가속), 그 이득을 포트폴리오 전체에 복리로 쌓습니다.
+  이미 Claude Code 에 같은 말을 반복하고 계실 겁니다. 돌려야 할 검사, 지켜야 할 규칙,
+  변경이 갖춰야 할 모양.<br>
+  <b>forge-harness 는 그걸 재사용 가능한 것으로 바꿉니다.</b> 저장소 안에 살면서 스스로 발화하는
+  스킬과 게이트와 에이전트로요.<br>
+  스킬은 일부러 범용으로 두고 쓰시는 동안 사례에 맞춰 벼립니다.
+  <b>같은 모양이 자꾸 돌아오면 그때 출하를 제안합니다</b> — 별도 스킬로, 또는 별도 하네스로.
 </p>
 
-<p align="center">
-  <b>품질이 지렛대이고, 속도는 그 결과입니다.</b> 모든 변경은 게이트를 통과해 제 값을 증명합니다 —<br>적대(adversarial) · 팬텀(phantom) · 회귀(regression) — 그리고 <i>그것</i>이 다음 변경을 더 빠르게 만듭니다.
-</p>
+---
+
+## 2분이면 됩니다 — 이 문서를 다 읽지 않으셔도 됩니다
+
+```bash
+claude plugin marketplace add https://github.com/chrono-meta/forge-harness.git
+claude plugin install -s user fh-meta@forge-harness
+git clone https://github.com/chrono-meta/forge-harness.git ~/projects/forge-harness
+cd ~/projects/forge-harness && claude
+```
+
+**그리고 `hi` 라고 입력하세요.** 번호가 붙은 메뉴가 뜨고 거기서부터는 도구가 안내합니다.
+문을 고르고 몇 가지에 답하면 설치 마법사까지 대신 돌려 줍니다. 이 아래는 필요할 때 찾아보는
+참고 자료이고, 시작하기 전에 읽어야 하는 숙제가 아닙니다.
+
+**증폭하는 것**: 시도 횟수입니다. 시행착오가 당신에게서 떨어져 나가 병렬로 돕니다.
+**증폭하지 않는 것**: 모델의 천장입니다. 하네스는 모델을 자기 천장까지 끌어올릴 뿐입니다.
+**확인하는 법**: 자기 등급을 공개합니다. 다섯 정체성을 정직하게,
+[릴리스](https://github.com/chrono-meta/forge-harness/releases)마다 적습니다. 초록이 아닌 칸은
+아직 없는 실제 실행이 무엇인지 이름으로 말합니다.
+
+---
 
 <p align="center">
-  <i>포크하세요. 이름을 바꾸세요. 당신의 것으로 만드세요.</i>
+  <img src="docs/pillars.svg" alt="FORK - ADAPT - COLLABORATE - EMPOWER" width="680">
 </p>
 
 <p align="center">
-  <img src="docs/pillars.svg" alt="FORK · ADAPT · COLLABORATE · EMPOWER" width="680">
+  <b>품질이 지렛대이고 속도는 그 결과입니다.</b> <i>포크하세요. 이름을 바꾸세요. 당신의 것으로 만드세요.</i><br>
+  <sub>도움이 됐다면 ⭐ 하나가 다른 분이 찾는 데 도움이 됩니다.</sub>
 </p>
 
 <p align="center">

package/README.md

@@ -18,24 +18,48 @@
 </p>
 
 <p align="center">
-  <sub>If this is useful, a ⭐ helps others find it.</sub>
+  <b>Ask it for things. When the asking repeats, it offers to build you the thing.</b>
 </p>
 
 <p align="center">
-  <b>Forge your Claude Code projects — pass them through, they come out faster.</b><br>
-  A practitioner's <b>meta-harness</b> — the galaxy your project harnesses live in.<br>It raises each project's <b>floor</b> (harness-ify the setup) and <b>ceiling</b> (accelerate the work), then compounds the gains across your whole portfolio.
+  You already tell Claude Code the same things over and over — the checks to run, the rules to hold,
+  the shape a change has to have.<br>
+  <b>forge-harness turns that into something reusable</b>: skills, gates and agents that live in your
+  repo and fire on their own.<br>
+  Its skills stay general on purpose and get shaped to your case as you go.
+  <b>When one shape keeps coming back, it offers to ship it</b> as its own skill, or its own harness.
 </p>
 
-<p align="center">
-  <b>Quality is the lever; speed is the result.</b> Every change earns its way through the gates —<br>adversarial · phantom · regression — and <i>that</i> is what makes the next change faster.
-</p>
+---
+
+## Try it in two minutes — you do not have to read this document
+
+```bash
+claude plugin marketplace add https://github.com/chrono-meta/forge-harness.git
+claude plugin install -s user fh-meta@forge-harness
+git clone https://github.com/chrono-meta/forge-harness.git ~/projects/forge-harness
+cd ~/projects/forge-harness && claude
+```
+
+**Then type `hi`.** A numbered menu appears and takes it from there — pick a door, answer a couple of
+questions, and it runs the install wizard for you. Everything below this line is reference for when you
+want it, not homework before you start.
+
+**What it amplifies** — the number of attempts; trial and error moves off you and runs in parallel.
+**What it does not** — the model's ceiling. A harness lifts a model to its own ceiling, not past it.
+**How you can check** — it grades itself in public, five identities, in every
+[release](https://github.com/chrono-meta/forge-harness/releases); the ones that are not green name
+the real run still missing.
+
+---
 
 <p align="center">
-  <i>Fork it. Rename it. Make it yours.</i>
+  <img src="docs/pillars.svg" alt="FORK - ADAPT - COLLABORATE - EMPOWER" width="680">
 </p>
 
 <p align="center">
-  <img src="docs/pillars.svg" alt="FORK · ADAPT · COLLABORATE · EMPOWER" width="680">
+  <b>Quality is the lever; speed is the result.</b> <i>Fork it. Rename it. Make it yours.</i><br>
+  <sub>If this is useful, a star helps others find it.</sub>
 </p>
 
 <p align="center">
@@ -59,7 +83,7 @@
 
 ---
 
-## Get started in 2 minutes
+## Requirements
 
 **Prerequisite**: Claude Code CLI — verify with `claude --version`
 
@@ -76,11 +100,8 @@ the single place a new machine can learn it. That is an improvement over nowhere
 python3 -m pip install --user pyyaml     # verify:  python3 -c 'import yaml; print(yaml.__version__)'
 ```
 
-Why this is called out rather than left implicit: a release once shipped green from a session whose
-`python3` happened to resolve to an **unrelated project's virtualenv** that had PyYAML, while the
-machine's own `python3` did not. The gate was never bypassed — it passed, and the pass simply was not
-portable. Every verdict from that gate now prints the interpreter and PyYAML version it used, so a
-green states what produced it instead of leaving the reader to assume.
+Every verdict from that gate prints the interpreter and PyYAML version it used, so a green
+states what produced it.
 
 </details>
 
@@ -141,6 +162,36 @@ cd ~/projects/{your-project} && claude
 
 ---
 
+## Two version numbers, and they measure different things
+
+This repo publishes **two counters**, deliberately. Conflating them is the single most common way to
+misread the project's status, so they are named here rather than only in the canon.
+
+| Counter | Where you see it | What it means |
+|---|---|---|
+| **Package version** (currently **2.7.0**) | npm, the plugin manifests, `git tag v2.x` | *what you install.* Ordinary release numbering: fixes → patch, new assets and gate lanes → minor, a capability **class** appearing or the thing being rebuilt → major |
+| **Identity-maturity release** (currently **identity-v0.4.0**) | the GitHub **Releases** page | *how far along the harness is.* `0.x` carries an incomplete-but-honest status **by design**; **the all-green ship is reserved for `identity-v1.0.0`** — every one of the five identities at 🟢, none 🔵/🟡/🔴 |
+
+🟥 **A high package number does not mean maturity.** `2.7.0` is not "ahead of" `identity-v0.4.0`; they are not on
+the same scale. The maturity track is deliberately allowed to sit at `0.x` while the package ships and
+improves, because the thing `0.x` refuses to do is **lie** — it says out loud that not every identity has
+cleared its bar yet, and each release names exactly which real run is still missing.
+
+⚠️ **Fixed, and the wart is left on the record**: the two counters used to share one `vX.Y.Z` git-tag
+namespace, and only the maturity track had GitHub *Release* objects — so the Releases page showed
+`v0.3.0` as "Latest" while the shipped package was `2.6.0`. Two layers under one name is a defect this
+project keeps finding in its own gates; here it was in its own version numbers. The maturity track now
+carries its own `identity-v*` prefix (first such release: `identity-v0.4.0`, 2026-08-21). 🟥 **Not** by also publishing the package
+track here — that was tried on 2026-08-21 and reverted the same hour: GitHub gives exactly **one**
+"Latest" badge, so two tracks on one page compete for it, and whichever holds it defines what the repo
+says it is. Putting the package number there pushed the maturity claim — the honest core — below it.
+**The Releases page carries the maturity track; what the package shipped is carried by
+[CHANGELOG](plugins/fh-meta/CHANGELOG.md) and the registry.** Existing tags are left alone — renaming them is an irreversible operation on a public surface, and the
+[Destructive-Op gate](knowledge/shared/harness-core/claude_md_gate_details.md) applies to us too.
+
+Full rules for what each grade requires:
+[`ship_readiness_gate.md`](knowledge/shared/harness-core/ship_readiness_gate.md).
+
 ## What it is
 
 forge-harness is structured as **two distinct layers**:
@@ -321,11 +372,26 @@ That is why the column that matters most below is *what it gets*:
 |---|---|---|---|
 | **ⓐ Different family** | the diff + the author's framing | the **implementation** is wrong | a reviewer from another model family (`auto-decorrelation`) |
 | **ⓑ Standpoint** | the diff + **the target harness's own canon** | **whether the rule you cited actually says that** | run the diff from that harness's own repo and rules ([`§7`](knowledge/shared/harness-core/field_verdict_crossfamily_gate.md)) |
-| **ⓒ Isolated grounding** | the sentences the author wrote + the tree as it stands now | the **claim** is wrong | someone who did not write it re-measures what it says |
+| **ⓒ Isolated grounding** | the sentences the author wrote — their claims *and* what they **declared before starting** — + the tree as it stands now | the **claim** is wrong · the delta does not match what was declared | someone who did not write it re-measures what it says; for the pre-declaration, a gate that reads the stated success definition back against the delta |
 | **ⓓ Third-party encounter** | the problem + **someone else's codebase** | **is this already solved** · where your change touches someone else's repo | look at the same problem in an unrelated third repo |
 | **ⓔ First real use** | one real target | the **way you are measuring** is wrong — the instrument's instrument | run it once against one real target and check the result by hand |
 | **ⓕ Revert and observe** | the tree with the wiring deleted | the **anchor** is wrong — the check is decorative | delete the thing it guards and confirm *that specific* check goes red |
 
+> **ⓒ widened on 2026-08-21, and how it widened is the more useful part.** Every commit marker in
+> this repo has been required since 2026-08-09 to carry the author's own pre-declaration — *what
+> counts as success* and *what I will not do* — written before designing. Measured with a control
+> that day: **nothing read it.** Zero lines of consuming code anywhere, while the sibling fields
+> were checked in 21 places; the gate spec did not even name it. On the real corpus, **37 of 98
+> markers carried no such line at all** — including a panel-reviewed one with 28 lanes and every
+> other field filled. The axes all looked *outward* (the diff, the target repo, prior art, the
+> artifact); none looked at the record's own mandatory field. A slot with no consumer always
+> reports "done", because presence is doing the judging.
+>
+> The fix was not a seventh axis. ⓒ already receives *the sentences the author wrote plus the tree
+> as it stands* — which is, word for word, what a pre-declaration check receives. Tense (declared
+> beforehand vs claimed afterwards) is a **posture**, like adversariality, not an axis. Minting a
+> new one would have repeated the exact error the blind reclassification above found.
+
 **You do not run all six every time, and that is the design** — do not multiply them, **choose**:
 
 ```
@@ -354,6 +420,37 @@ with tools**, the boundary blurs — an outside judgment held that "the store is
 Conversely, "a rule another repo retired long ago" **cannot be fetched by any tool** — there is no reason
 to have access to that project's review history in the first place. That is where ⓓ remains.
 
+**Where a rule lives — and why the always-loaded layer does not have to grow forever.**
+
+A harness learns by writing rules down. The obvious place is the always-loaded file every session
+reads, and that file only ever gets longer. Left there, the reasoning ends in a corner: *a harness
+that keeps learning keeps getting more expensive to start.*
+
+It does not, because a rule has **three possible seats**, and the right one is decided by **when the
+rule has to fire**:
+
+| Seat | Fires | Costs | Fits |
+|---|---|---|---|
+| **Always-loaded** | before you act | every session, every turn | rules whose trigger is an *intention* — tone, "don't normalize the unfamiliar", "prove the instrument works here". Nothing can hook an intention, so salience is the only layer |
+| **The gate's own error message** | at the moment you act | **nothing** | rules whose trigger is an *action*. The message that blocks you also teaches the form: `Write, before the design: success = «…». never = «…».` |
+| **The hook** | after you act | nothing | properties of a record — present · typed · attributable · non-vacuous |
+
+The middle seat is the one that usually goes unused, and it is free. It is
+[gate-locality](knowledge/shared/harness-core/gate_locality_principle.md) applied to salience: the actor reads it exactly where the
+action happens, so it does not have to be carried all session to be there when needed.
+
+🟥 **It is a third layer, not a replacement — and the honest limit is that it only fires on failure.**
+Someone who gets it right never sees it. So mechanizing a rule does **not** shrink the resident layer:
+measured on the very change described above, the machine grew by 480 lines and the always-loaded prose
+by **zero**, and that is correct. The prose has to reach the author *before* they design; the hook
+catches its absence *after*. A backstop cannot substitute for salience that must fire earlier.
+
+⚠️ And the threshold that would tell you the resident layer is "too big" is, in this repo, **not
+grounded** — the numbers in our own doctor skill were introduced without a single line justifying the
+cutpoints, and one of them was set to a value the target already exceeded on the day it landed. We are
+re-deriving them rather than trimming toward a number nobody can defend. Cutting resident text toward
+an unjustified target buys fail-open with the savings.
+
 > 🟥 **Limits to read before citing this**: the six-axis table is **n=1** (one artifact · one session ·
 > one author). Whether the axes' non-overlap is structural or an accident of that day is **unmeasured**.
 > And when the author's self-scoring was stripped out — 16 findings handed, **with their provenance

package/README.zh.md

@@ -18,24 +18,44 @@
 </p>
 
 <p align="center">
-  <sub>如果这对你有用，⭐ 一下能帮助更多人发现它。</sub>
+  <b>有需要就交给它。同一个请求反复出现时，它会主动提议把那件事本身做出来。</b>
 </p>
 
 <p align="center">
-  <b>锻造你的 Claude Code 项目 —— 让它通过，它会更快出炉。</b><br>
-  一个实践者的 <b>元框架 (meta-harness)</b> —— 你的项目框架们所栖居的星系。<br>它抬高每个项目的 <b>下限 (floor)</b>（把设置框架化）和 <b>上限 (ceiling)</b>（加速工作），再把这些收益在你的整个项目组合中复利累积。
+  你大概已经在对 Claude Code 反复说同样的话：要跑的检查、要守的规则、一次变更该有的样子。<br>
+  <b>forge-harness 把这些变成可复用的东西</b>：住在你仓库里、会自己触发的技能、闸门和 agent。<br>
+  技能刻意保持通用，在使用过程中按你的场景当场锻造。
+  <b>当同一种形状反复回来，它就提议出货</b> —— 作为独立技能，或独立框架。
 </p>
 
-<p align="center">
-  <b>品质是杠杆，速度是结果。</b> 每一次变更都要挣得通过门禁的资格 ——<br>对抗 (adversarial) · 幽灵 (phantom) · 回归 (regression) —— 而 <i>正是这一点</i>让下一次变更更快。
-</p>
+---
+
+## 两分钟就能试 —— 你不必读完这份文档
+
+```bash
+claude plugin marketplace add https://github.com/chrono-meta/forge-harness.git
+claude plugin install -s user fh-meta@forge-harness
+git clone https://github.com/chrono-meta/forge-harness.git ~/projects/forge-harness
+cd ~/projects/forge-harness && claude
+```
+
+**然后输入 `hi`。** 会出现一个带编号的菜单，之后由工具引导你：选一个入口，回答几个问题，
+它会替你运行安装向导。这条线以下是需要时再查的参考资料，而不是开始前的作业。
+
+**它放大什么**：尝试的次数。试错从你身上移开，并行运行。
+**它不放大什么**：模型的天花板。框架只把模型抬到它自己的天花板，不会更高。
+**如何验证**：它公开自己的评级。五个身份，如实记录在每一次
+[发布](https://github.com/chrono-meta/forge-harness/releases)中。未变绿的那些，会指名说出还缺哪一次真实运行。
+
+---
 
 <p align="center">
-  <i>Fork 它。改名。让它成为你的。</i>
+  <img src="docs/pillars.svg" alt="FORK - ADAPT - COLLABORATE - EMPOWER" width="680">
 </p>
 
 <p align="center">
-  <img src="docs/pillars.svg" alt="FORK · ADAPT · COLLABORATE · EMPOWER" width="680">
+  <b>质量是杠杆，速度是结果。</b> <i>Fork 它。改名。让它成为你的。</i><br>
+  <sub>如果这对你有用，⭐ 一下能帮助更多人发现它。</sub>
 </p>
 
 <p align="center">

package/knowledge/shared/harness-core/field_verdict_crossfamily_gate.md

@@ -469,11 +469,25 @@ own. `tier2b` is the honest reachable rung for that pairing; do not inflate a `t
 `tier3`, and do not undersell it to `tier2` either — it is a distinct, real, if operator-correlated,
 data point.
 
-**Mechanization status — `standpoint:` is prose-only today, and this must not be read as more than
-that.** `crossfamily:`'s degrade triad is hard-blocked at commit (`templates/.git-hooks/pre-commit`,
-`scripts/test_marker_crossfamily_lanes.sh` — grep-verified: ~20 crossfamily references in the hook,
-validated fixtures). `standpoint:` has **no value-enum validation and no fixture suite** — nothing stops an author from
-writing `not-applicable` with a thin justification, and no marker-shape check catches a wrong value.
+**Mechanization status — 🟥 CORRECTED 2026-08-20.** This paragraph used to say `standpoint:` was
+"prose-only today" with "**no value-enum validation and no fixture suite**". **Both halves are false**
+and had been for a while: `validate_standpoint_leg()` lives in `templates/.git-hooks/pre-commit` and
+`scripts/test_marker_standpoint_lanes.sh` is wired through `scripts/selfcheck.sh`. Measured by varying
+one variable at a time — `banana(qasp)` → **blocked (enum)** · `tier2` without parens → **blocked
+(enum)** · `tier2(qasp)` with no execution grounds → **passes with a warning** · with grounds →
+**passes**.
+
+**What is actually true, stated at the right width**: the enum IS closed and enforced; the `tier2`+
+execution grounds are **advisory** (a thin `tier2` records and warns, it does not block); and nothing
+checks whether the recorded value is *true*. The old sentence collapsed all three into "no validation",
+which suppresses use of a control that exists — the quietest kind of drift, because it reads as
+honest modesty.
+
+🟥 **This same false claim stood in FIVE places** (`CLAUDE.md` ×2, `.claude/rules/fh_4axis_gate.md`,
+`AGENTS.md`, here). The first repair fixed one; the second fixed three and still missed this file and
+over-stated the fix. Both misses were surfaced by a **different-family reviewer**, not by a lane and
+not by the author — there is still no check that compares a rule's self-description against the
+machinery it describes.
 
 🟥 **Two sentences that stood here were STALE and are corrected (2026-08-17, re-measured — a
 cross-family reviewer flagged the second, the first fell out of checking it).** They read