fix(verification,testing): stop routing live-app verification through the user-invoked-only /verify - #1878
Conversation
…r-invoked-only /verify RA-10 alleged that the `≥2.1.145` gate on the bundled `/verify` + `/run` note in `plugins/verification/skills/confirm/SKILL.md` was stale as an availability statement, because corpus entry verifblog OQ-01-1 records that Claude Code v2.1.215 made bundled `/verify` and `/code-review` user-invoked only. That allegation is refused. A live fetch of https://code.claude.com/docs/en/skills.md on 2026-08-02 shows the two facts sit on orthogonal axes: line 45 still states "All three skills require Claude Code v2.1.145 or later", so the floor is current, while line 25 restricts who may invoke — "others, including `/verify` and `/code-review`, run only when you invoke them". v2.1.215 changed the invoker, not the version floor. The `2.1.145` floor value is therefore unchanged wherever it appears — no site's version was altered. Two of the four pre-existing mentions outside `.work/` sit on lines this commit rewrote for the invocation fix, and the two new changelog entries restate the floor, so the branch carries six mentions against origin/main's four. The investigation did surface a real, adjacent defect on the same lines: these skills instruct the *agent* to invoke `/verify`, a surface the model may no longer invoke. Six delegation sites named it across the two plugins; the four that told the skill to reach for it are corrected, and the two that remain now frame `/verify` as a suggestion for the user to run. `/run` is untouched — the v2.1.215 sentence names neither it nor `run-skill-generator`, so it stays the agent-invocable supplementary path. The `/code-review` half of the same upstream change is deliberately out of scope: its 23 references live in the `review` plugin and are predominantly descriptive boundary prose or already user-gated, a distinct population needing its own assessment. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_015uHDhn2Jfd1N6M1S3AsMSf
…d-only /verify The preceding commit stopped `/verification:confirm` from delegating live-app verification to the bundled `/verify`, because Claude Code v2.1.215 made `/verify` and `/code-review` run only when the user invokes them. Four prose sites moved; the skill's graded rubric did not. `plugins/verification/skills/confirm/evals/evals.json` eval 9 (`live-app-delegates-to-bundled-with-fallback`) still encoded the removed delegation as its pass condition: `expected_output` required "delegates to /testing:run-e2e (primary for orchestrated apps) or bundled /verify + bundled /run", and its second expectation required "bundled /verify + /run as supplementary". Against SKILL.md:112 — "so suggest the user run it rather than delegating to it — this skill cannot invoke it" — a grader reading both marks the corrected skill FAIL. The branch carried a skill contradicting its own rubric; no released version did, since both halves land in unreleased 0.3.6. Eval 9's `expected_output` and expectations now state what the skill does: `/testing:run-e2e` primary for orchestrated apps, the bundled `/run` supplementary and agent-invocable, the bundled `/verify` surfaced as a suggestion for the user rather than invoked, and the graceful fallback keyed to `/run` alone rather than the ambiguous "bundled-tool". The case `id`, `name`, and `prompt` are unchanged — `/run` delegation genuinely remains, so the name still describes the case and nothing else in the repo cites it. Population re-enumerated rather than inherited. `git grep -n -I -E '/verify([^a-zA-Z0-9_:-]|$)' -- .` returns 49 lines: 34 outside the `verification` and `testing` plugins are unrelated (docs-hygiene's `/verify` rename examples, `actions/verify.md` path refs, `fix/verify/report` prose), and of the 15 in-scope lines every one but these two was already corrected or is an immutable shipped-version CHANGELOG record (testing 0.2.x line 58, verification lines 68/137/170). Two token-free sweeps closed the residual gap: `git grep -n -I 'bundled-skills' -- .` returns 3 lines, all already-corrected sites, and `git grep -n -I 'run-e2e' -- . ':!plugins/testing' ':!plugins/verification'` returns 5 lines, none naming the bundled `/verify`. No model-graded eval runner exists in this repo, so the regression was proven by entailment against the shipped text rather than a run. The executable gates CI runs were executed: `check-jsonschema --schemafile plugins/skill-quality/reference/evals.schema.json` exits 0 on the edited file, `scripts/check-changed-skills.sh origin/main` reports both touched skills PASS with 0 errors, and `scripts/check-cross-plugin-source-drift.sh` exits 0 without implicating either changed file. The orphaned-fixtures gate is structurally unaffected — the plugin has no `evals/fixtures/` directory and eval 9's `files` array is empty and untouched. The 0.3.6 CHANGELOG entry gains one clause noting the rubric moved with the behavior; the version bump already on this branch is not re-bumped. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_015uHDhn2Jfd1N6M1S3AsMSf
|
Claude finished @kyle-sexton's task in 3m 1s —— View job Review: fix(verification,testing) — stop routing live-app verification through the user-invoked-only
|
|
Claude finished @kyle-sexton's task in 56s —— View job Security review complete
No security issues found. This PR touches only prose/documentation content and version metadata across 8 files:
No GitHub Actions workflows, hook scripts, |
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 2457dee4d1
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
|
Claude has reviewed this PR 3 times. The lane skips further automatic reviews after 5; deleting this comment resets the count. |
…to v2.1.215 The branch's correction overshot. `plugins/verification/skills/confirm/SKILL.md` said `/verify` "is user-invoked only ... this skill cannot invoke it" with no version attached, and `plugins/testing/skills/run-e2e/SKILL.md` said "[user-invoked only]". Both statements are true for current clients and false for 2.1.145-2.1.214, where `/verify` exists and Claude may still run it. The originating finding conflated availability with invocability; an unscoped restriction re-introduces the same conflation from the opposite side. Upstream is explicit and version-scoped. A raw `curl` of https://code.claude.com/docs/en/skills.md on 2026-08-02 returned 73907 bytes / 933 lines; line 25 reads "others, including `/verify` and `/code-review`, run only when you invoke them" followed by "Before v2.1.215, Claude could also run `/verify` and `/code-review` on its own", and line 45 reads "All three skills require Claude Code v2.1.145 or later". Two axes, two versions. An unscoped statement would be correct only behind a declared support floor, and this repository declares none. `grep -rniE 'minimum (supported )?(claude code )?version|min(imum)? claude code|requires claude code|support (floor|window)| supported versions|oldest supported'` over tracked `.md`/`.json`/`.yml` returns 7 lines, none of which is a repo-wide Claude Code floor: two are the per-claim `v2.1.206` gate on `/doctor`'s trim, one is this very line 112, one is lint criteria prose, one is Kindle tooling, one is a decision-matrix bullet, and `SECURITY.md` "Supported Versions" governs plugin versions, not client versions. No `plugin.json` or `marketplace.json` carries an `engines` / `minClaudeCodeVersion` / `requiredVersion` field (`grep -rlE` returns zero files), and `docs/PLUGIN-PHILOSOPHY.md` / `docs/MIGRATION-PLAYBOOK.md` state none. An undeclared floor is not a floor, so the claim gets the version. The repo's own convention agrees: `plugins/claude-config/skills/ audit-instructions/reference/criteria.md:328` — "Where the behavior is version-gated, carry the minimum version with the claim" — and the neighbouring `>=2.1.145` availability note on the same lines is exactly that shape. That note is untouched and still its own axis; the two are now named as separate axes so the correction cannot be misread as re-merging them. The *instruction* stays uniform on purpose and is not branched on the client's version: suggesting the user run `/verify` is correct on every version it exists on, so no runtime version probe is warranted for a closed window. The behavior therefore does not regress on pre-2.1.215 clients - it simply does not autonomously delegate there either, which the changed prose now states rather than justifying with a false absolute. Population enumerated, not assumed. `grep -rniE 'user-invoked|user invoked| agent-invocable|agent invocable|model-invocable|cannot invoke|cannot reach|only when you invoke'` over tracked `.md`/`.json` returns 113 lines repo-wide, dominated by unrelated `cannot reach` changelog boilerplate; discriminating on the bundled `/verify` leaves 6 sites across the two plugins. Three stated the restriction unscoped and are corrected here (`confirm/SKILL.md:112`, `run-e2e/SKILL.md:61`, and eval 9's `expected_output` plus its second expectation in `confirm/evals/evals.json`). The two unreleased CHANGELOG entries (verification 0.3.6, testing 0.3.3) already named v2.1.215 in-sentence and gain the scope explicitly; no shipped-version entry was touched. `README.md` made no invocability claim after the earlier commit removed `/verify` from the fallback sentence, so it needed nothing. The `/code-review` half of the same upstream sentence was swept rather than inherited: `grep -rn '/code-review'` returns 40 lines, all in the `review` plugin, `playbooks` vendor text, and design docs. None asserts an invocability restriction in either direction - the `review` surfaces already gate `--fix` / `--comment` on explicit user opt-in - so there is no parallel overreach to scope, and the second commit's deliberate scope-out stands. Gates run on the edited files: `markdownlint-cli2` reports 0 errors across all four markdown files, `check-jsonschema --schemafile plugins/skill-quality/reference/evals.schema.json` exits 0 on the eval file, `scripts/check-changed-skills.sh origin/main` reports both touched skills PASS with 0 errors (the one warning each is pre-existing description phrasing), and `scripts/check-cross-plugin-source-drift.sh` exits 0 without implicating any changed file. No version is re-bumped; 0.3.6 and 0.3.3 remain unreleased. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_015uHDhn2Jfd1N6M1S3AsMSf
|
Claude finished @kyle-sexton's task in 1m 42s —— View job Review: fix(verification,testing) — stop routing live-app verification through the user-invoked-only
|
|
Claude finished @kyle-sexton's task in 51s —— View job Security review complete
No security issues found. All 8 changed files in this PR are non-executable content — no code paths, no workflow files, no scripts:
No GitHub Actions workflow files, hook scripts, |
Two defects in the preceding commit, both caught by independent review. `plugins/testing/skills/run-e2e/SKILL.md:61` read "user-invoked only from v2.1.215 - earlier versions in that window let Claude run it too". No window had been named: the only prior version on the line is the open-ended `>=2.1.145` floor. The phrase resolved to nothing, in the exact file the review thread names as a suspected sibling site. It now matches the wording its `confirm/SKILL.md:112` sibling already used - "before v2.1.215 Claude could also run it on its own" - and the trailing clause names the window it means (`>=2.1.145`) instead of gesturing at an undefined one. No fact changed; the sentence now states the one it always intended. The preceding commit message also cited "40 lines" for the `/code-review` sweep. That number was never produced by a command. Piped through `wc -l`, `grep -rn '/code-review' --include="*.md" --include="*.json" . | grep -v '^./.git/'` returns 48, distributed `plugins/review` 31, `plugins/playbooks` 9, `docs` 5, and one line each in `plugins/verification`, `plugins/source-control`, and `plugins/docs-hygiene`. The qualitative finding that number supported is unchanged and still holds: none of the 48 asserts an invocability restriction in either direction - the `review` surfaces gate `--fix` / `--comment` on explicit user opt-in, the `playbooks` lines are vendored upstream tip text, and the rest are boundary or naming prose - so `/code-review` carries no parallel overreach to version-scope. The earlier commit's separate "23 references" figure counted the `review` plugin alone at an earlier tree state and is not comparable; 48 is the repo-wide count at this commit. Recorded here rather than by amending, because the branch has an open PR and its review record must stay intact. `markdownlint-cli2` reports 0 errors on the edited file and `scripts/check-changed-skills.sh origin/main` reports all 5 touched skills PASS with 0 errors. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_015uHDhn2Jfd1N6M1S3AsMSf
|
Claude finished @kyle-sexton's task in 1m 17s —— View job Review: fix(verification,testing) — stop routing live-app verification through the user-invoked-only
|
|
Claude finished @kyle-sexton's task in 53s —— View job Security review complete
No security issues found. All 8 changed files are non-executable content — no workflow files, scripts, or code paths are part of this diff:
No GitHub Actions workflow files, hook scripts, |
|
Accepted on the facts, with one deliberate partial. You are right that the shipped statement overreached; I took option 1 — version-scope it — because the repo has no support floor that would license the unscoped form. Upstream, fetched raw. The floor search was a null, so option 2 was not available. A pattern sweep for a repo-wide client floor returns 7 lines, none of which is one: two are the per-claim The repo's own convention says the same thing directly — What changed. The Where I declined, and why. Your title asks to preserve delegation on pre-2.1.215 clients. I preserved the statement's accuracy but left the instruction uniform rather than branching on client version. That is a decision, not an omission: suggesting the user run Sibling swept. |
…nnel extraction artifacts (#1879) Second batch of evidence-forced docpage-digest amendments. `knowledge` 0.10.15 -> 0.10.16. Two rules ship; six items are deliberately held, each with its reason stated. ## What ships - **PA-Y — the near-miss disclosure rule, written into the profile.** It previously lived only inside Ruling 1 Decision A and was re-derived per unit, which is why undisclosed near-misses are the largest MINOR class in the slice that measured it. A harness page covering the same subject without stating the claim's specific rule is a near-miss: the tag survives, and the row must disclose it by name and line. - **PA-Z — blog-channel extraction artifacts** (collapsed H1, split reading-time) with reconstruction guidance. Its deferral condition was met: a second `claude.com/blog` run reproduced both exactly, re-verified byte-exact in both slices. The guidance leads with the canonical URL slug and states what a slug **loses** — word boundaries recover, punctuation and casing do not — so a slug-recovered title is labelled reconstructed. - **PA-AG — verified and closed, no edit.** The `cc-applicable`/`mixed` boundary rule already in the profile matches what its source asked for. ## What is held, and why — two different reasons **Five held because the campaign's judgment file holds them.** PA-I (J-6), PA-Q (J-7), PA-N (J-8), PA-M (J-12) and PA-AD (J-14) were shipped in the first draft of this branch and then withdrawn. The triage marks them `evidence-forced`; `PHASE2-JUDGMENT-AMENDMENTS` writes them up as judgment calls under a header stating "None applied." Two records disagreed, and the judgment file governs on its own reasoning: - J-14 says outright it is "held here only because no defect was demonstrated, so the bar for the applied subset is not met." - J-7 says it "should be decided **after** Decision-A ordering" — an escalated, owner-reserved question. Shipping it would have pre-empted a decision the owner kept for himself. They are now full rows in the owner-decision artifact, each restating the triage-vs-judgment conflict rather than resolving it silently. Between the two readings they would otherwise have landed in neither place. **Three held because they change untracked instruments.** PA-B's second half, PA-P and PA-AK all modify scripts under the memory-tier `.work/` root, which is never committed and is not part of the shipped plugin. ## Verification Independently verified twice. The first round FAILED on the five judgment-held rules. The second confirmed the split is clean: `SKILL.md` is byte-identical to its 0.10.15 state, grep finds no residue of the withdrawn rules, and PA-Y reads as a whole rule rather than half of the graded pair it was originally merged into. The implementer also ran a second pass over the whole *file* rather than the diff, and caught three defects in its own first commit — including a new rule that contradicted one shipped in #1875, which is exactly the two-rules-for-one-boundary drift this pipeline exists to prevent. No linked issue ## Related - Follows #1875 (`knowledge` 0.10.15), which shipped the first evidence-forced batch. - Phase 3a self-alignment merged in #1876, #1877, #1878; this repo conforms before any of it ships outward. --------- Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
…urfaces (#1880) Applies the evidence-forced half of the RA-9 extend-or-cite pass to instruction-audit catalog row **I10** (reasoning-echo). `claude-config` 0.17.0 -> 0.18.0; `criteria.md` 1.4.0 -> 1.5.0. ## What this does NOT do, and why that is the point **I10 keeps `Model scope: fable-5`. It is not promoted, and `mythos-5` is not added.** The campaign triage ranked this promotion its **#1 item**, on the theory that the extended-thinking platform page is a second, model-independent source that would satisfy I10's promotion gate. It is not, and the page's own structure is what settles it: - `:874` — the H2 reads "Thinking output on Claude Fable 5 **and Claude Mythos 5**" - `:876` — names **both** models for the adjacent raw-chain-of-thought property - `:887` — names **only** Claude Fable 5 for `stop_details.category: "reasoning_extraction"` Eleven lines apart, in one section. The page had a sentence-adjacent opportunity to widen the refusal and declined it. That is deliberate scoping, not loose phrasing. Adding `mythos-5` would fabricate scope from a claim about a different property — exactly what the catalog's own model-scoping block warns against. That reasoning is now recorded **in the row itself**, so it is not re-litigated a fourth time. ## What it does - **A corroborating second Source** — the extended-thinking page, independent of the Fable 5 model guide already cited. - **A concretized Remediate line.** It previously said, abstractly, to read structured `thinking` blocks or use a send-to-user tool. It now names the actual surfaces: `Ctrl+O` verbose mode and `showThinkingSummaries: true`, and `display: "summarized"`. **Each surface is attributed to the page that actually states it.** The pass this work came from asserted all three were on the thinking page; two of them are not — `Ctrl+O` and `showThinkingSummaries` are stated at `model-config:532`, and only `display: "summarized"` is on the thinking page. Both pages are therefore added to `## Sources`, because `criteria.md` carries its own invariant that "the trigger set is the source set — naming a subset would leave the harness-behavior rows depending on pages nothing watches." ## Verification Independently verified by a second model, with the implementer's rationale withheld. Both pages re-fetched raw (`model-config.md` 83,644 bytes; `thinking.md` 52,769 bytes, byte-identical to the frozen snapshot), all three surface names confirmed verbatim, and the promotion-gate facts re-confirmed at both snapshot and live bytes. **Self-fire check.** Because a catalog row that fires on this repo would break the campaign's own rule against shipping a consumer check we fail: the scanner was run against both trees. I10 candidates in `criteria.md` go 2 -> 4, and all four are inert — `criteria.md` is a skill reference file, outside the audited population (`CLAUDE.md` / `rules/` / `skills/` / `agents/` / `output-styles/` under the user and project roots). No new class of self-hit is introduced. Detection is untouched: `instruction-scan.test.sh` reports 46/46, and this edit changes Source and Remediate only. ## Held back deliberately The **I8-b promotion**, whose gate genuinely IS met by verbatim two-guide convergence, is not here. Promoting it makes the row fire on this repo's own `plugins/review/context/severity.md` — the same work as triage row RA-2, an open owner decision. Shipping the promotion first would make the next audit run flag this repository. No linked issue ## Related - Phase 3b of the doc-corpus campaign; Phase 3a self-alignment merged in #1875, #1876, #1877, #1878, #1879. - Source pass: `RA9-EXTEND-OR-CITE-PASS-2026-08-02.md` (23 rows adjudicated; the triage had checked 6). Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
…y.md enumerable tier criteria (#1881) Promotes instruction-audit catalog row **I8-b** (conservative-reporting detection) to unscoped and gives `plugins/review/context/severity.md` enumerable tier criteria. `claude-config` 0.18.0 -> 0.19.0; `review` 0.15.5 -> 0.16.0; `criteria.md` 1.5.0 -> 1.6.0. ## The promotion I8-b's gate is MET on its second arm (multiple model guides converge): the Sonnet 5 prompting guide states all three trigger phrases verbatim in one sentence (`source.md:140`, "Code review harnesses"), converging with the Opus 5 guide. Annotated in row I7's met-gate precedent form; the Sonnet 5 guide URL added to `## Sources` per the trigger-set-is-the-source-set invariant. **Source attribution corrected while citing:** "don't nitpick" appears nowhere in the Opus 5 guide (`grep -cin nitpick` -> 0); that guide states only the other two phrases (its line 20). The Sonnet 5 guide is the phrase's only cited home. This strengthens the convergence gate — three phrases now each attributed to a page that actually contains them. ## #1880's "Held back deliberately" premise was wrong That PR deferred this promotion because unscoping I8-b would allegedly make it fire on `severity.md`. It does not, on three independently verified grounds: 1. **Zero scanner candidates** — the real scanner over `severity.md` emits one I6 row and no I8-b; the I8-b ERE greps 0 on both trees. 2. **I8-b's own carve-out** (`criteria.md:252-255`) excludes "severity-based routing where everything is still reported somewhere" — severity.md classifies findings and withholds none. 3. **Outside the audited population** — SKILL.md Phase A inventories CLAUDE.md / rules/ / skills/ / agents/ / output-styles/ under user and project roots plus hook text; `plugins/review/context/` is none of those (same result #1880 recorded for `criteria.md` itself). The promotion could have shipped alone. The severity.md work ships here anyway, **re-founded on its own source**: nine lines below the three-phrase line, the same Sonnet 5 guide says to "be concrete about where the bar is rather than using qualitative terms like `important`" (`source.md:150`) — and `important` is one of severity.md's own tier names. That is triage row RA-2, a distinct claim from the one I8-b cites. ## The severity rewrite Each tier now carries a decidable test instead of a qualitative label; no finding changes tier. Guards added where the criterion-stating change could have silently re-tiered: - The P1-P5 fold explicitly takes precedence for P-scored findings (otherwise every P3 security finding would have read into the new CRITICAL test). - CRITICAL's subsequent-change limb reads "otherwise-correct change", so cascade architecture violations (break a *correct* future change) stay CRITICAL while code duplication (bites only through an *incomplete* future edit) stays IMPORTANT. All eleven tier examples adjudicated against the new tests — twice, independently. ## Verification Independently verified by a second model with the implementer's rationale withheld, across two rounds. Scanner counts replayed from git refs both rounds: origin/main 23 rows / 6 files, all fenced; working tree 28 / 7, every addition this branch's own quoting. Planted-positive check: a constructed three-phrase file emitted I8-b on all three lines in the same invocation where severity.md emitted zero. Scanner-vs-git-grep equivalence proven on identical row sets. `instruction-scan.test.sh` 46/46; markdownlint 0 errors. **Known citation defect in an immutable commit message:** `3603c8a9b8` says "eight lines later (`source.md:149`)". Both figures are wrong — the correct citation is `source.md:150`, ten lines after the three-phrase line at `source.md:140`. No tracked file carries the wrong number; recorded here rather than rewriting pushed history. No linked issue ## Related - Phase 3b of the doc-corpus campaign; Phase 3a merged in #1875, #1876, #1877, #1878, #1879, #1880. - Reverses #1880's "Held back deliberately" reasoning with evidence (above). - Triage rows: RA-2 (severity criteria), RA-9/Q6 (I8-b promotion), owner decision Q9 (bundling). --------- Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
1 participant
Repo-alignment finding RA-10 from the doc-corpus campaign's Phase 3a. The finding as filed was refused; a different, real defect was found underneath it and fixed.
The refusal
RA-10 claimed the
>=2.1.145gate on the bundled/verifynote was stale. It is not. Livecurlof the raw skills reference (HTTP 200, 73,907 bytes, 933 lines — non-truncated, and the length is recorded because a rendered fetch of a long page can only manufacture a false absence) still states: "All three skills require Claude Code v2.1.145 or later" for/run,/verify,/run-skill-generator.The triage conflated two orthogonal axes.
2.1.145is the availability floor.2.1.215changed the invoker: "others, including/verifyand/code-review, run only when you invoke them... Before v2.1.215, Claude could also run/verifyand/code-reviewon its own." Nothing about an invocation restriction makes a version floor stale.The real defect, in the opposite direction
The triage asked whether the repo claims auto-invocation, and correctly found it does not. But the repo did something stronger: it instructed the agent to invoke
/verify— a surface the model may no longer invoke.Four live assertion sites now state the opposite:
/verifyis surfaced as a suggestion for the user to run, never delegated to./runis deliberately untouched — upstream restricts only/verifyand/code-review, and/runsurvives as supplementary and agent-invocable at all nine live sites.The eval gap
The first pass left
confirm/evals/evals.jsonencoding the removed delegation, so the skill would have shipped failing its own graded rubric. Caught by an independent verifier on population completeness — the campaign's most repeated error. Eval 9 is now entailed clause-by-clause by the shipped skill text instead of contradicting it, and its id/name/prompt are unchanged because/rundelegation genuinely remains.Changelog entries for already-shipped versions are deliberately left as written: a released entry is an immutable record of what that version did.
Verification
Two independent verification rounds, rationale withheld. The second re-enumerated from scratch rather than inheriting the first's list: 50
/verifylines repo-wide, 34 unrelated, 12 changelog text, 4 live assertion sites — all four corrected, zero remaining. Theagent-invocablewording now shipping inside a graded rubric was grounded empirically (this session's skills listing exposesrunwhile bundledverifyandcode-revieware absent) rather than on upstream's non-exhaustive "including".CI gates were run directly rather than trusted from the commit message, including the markdownlint pass that the skill-check harness had skipped.
No linked issue
Related