docs(ai-slop): record Ineffective indicators negative finding (0.3.2)

[kyle-sexton]

Closes #3032

Summary

ai-slop's catalog cited Wikipedia's Signs of AI writing but recorded Ineffective indicators (and Comment-specific indicators) as a fetch gap. That left a guardrail question open: if a shipped rule appeared in the section the page's own editors call unreliable, we would be shipping a tell our source rejects. A 2026-08-21 retrieval answers it: none of the 15 shipped script rules appear there, so no rule is dropped or re-scoped.

Fix

• Closed the fetch gap in plugins/ai-slop/skills/audit/reference/catalog.md against two revisions of https://en.wikipedia.org/wiki/Wikipedia:Signs_of_AI_writing:
  • Catalog pin 1369699198 (2026-08-16), MediaWiki parse sections 80 / 62 / 29.
  • Live page revision 1370403579 (2026-08-20T23:13:41Z), retrieved 2026-08-21.
• Quoted the eight ineffective indicators (perfect grammar; mixed registers; "bland"/"robotic" prose; undifferentiated formal prose; transition words in isolation; unsourced content; bizarre wikitext; correct wikitext). Same list on both revisions.
• Recorded that rule-em-dash and rule-rule-of-three remain valid source-page signs (Style / Language and grammar). The em-dash "in combination with other indicators" sentence is a corroboration qualifier, not an Ineffective indicators listing; zero-tolerance stays the house-style default.
• Comment-specific indicators retrieved in the same visit: Wikipedia talk-page scope, overlapping existing recorded-only edit-summary entries. No new slugs.
• Bumped ai-slop 0.3.1 → 0.3.2. Detector, emitter, evals, and rewrite guide are untouched.

Test plan

• npx markdownlint-cli2 on the two edited markdown files: 0 issues.
• python3 parse of plugin.json: version 0.3.2.
• bash plugins/ai-slop/skills/audit/scripts/detect.test.sh: all 92 cases passed (roster still exactly 15 script rules).
• scripts/check-changelog-parity.sh --check, --check-bump origin/main, --check-preserved origin/main: pass.
• scripts/check-changed-skills.sh origin/main: PASS, 0 errors.

Verification

Same commands and outcomes as the Test plan. This is a catalog/docs close of a fetch gap, not a detector change; the unit suite is the regression that the 15-rule roster did not move.

Related

• ai-slop: check whether any shipped rule appears under Wikipedia's "Ineffective indicators" #3032 — the fetch-gap / guardrail question this closes.
• feat(ai-slop): unslop-inspired catalog additions, plus the corrections and dogfood config that make them true (0.2.0) #3031 — the contract slice this issue graduated from.

[claude]

Claude finished @kyle-sexton's task in 2m 23s —— View job

---

Claude finished the code review

• Gather context (PR diff, description, changed files)
• Run /review:code-review per custom instructions
• Post findings as inline comments (if any)
• Post summary comment

Review scope

Docs-only change: plugins/ai-slop/skills/audit/reference/catalog.md, plugins/ai-slop/CHANGELOG.md, plugins/ai-slop/.claude-plugin/plugin.json. No detector, emitter, eval, or rewrite-guide code touched. No security-lane surface (no auth/data-handling/injection paths in this diff), consistent with REVIEW.md's split.

Findings

None that a careful reviewer would block or flag. Specifically checked and confirmed consistent:

• Roster-overlap claim (catalog.md line 578, "no shipped rule appears here"): verified independently by counting - v1: script tags in the file — 15, matching the claimed roster size. Cross-checked all eight quoted "Ineffective indicators" against the 15 script rules by category (em dash → ## Style; rule of three → ## Language and grammar; unsourced content vs. rule-utm-params in ## Citations; bizarre/correct wikitext vs. ## Markup's script rule rule-llm-citation-artifacts) — no overlap found in any case, matching the PR's verdict.
• rule-em-dash addition (catalog.md lines 274–281): the claim that the source page's em-dash sign lives under Style (a valid sign, not Ineffective) matches the file's own section structure (## Style at line 221 contains rule-em-dash).
• Comment-specific indicators section (catalog.md lines 443–459): claimed overlap with rule-canned-policy-assurance / rule-verbose-edit-summaries checks out against the ## Edit summaries section (lines 461–500), which are indeed both recorded-only/wikipedia-specific.
• CHANGELOG.md entry format (## [0.3.2], no date) matches the existing ## [0.3.1] convention in the same file.
• Version bump: plugin.json 0.3.1 → 0.3.2 matches the new CHANGELOG heading.

Not independently re-verified (out of scope for this lane / no tool access): the exact wording of the live Wikipedia quotes against the actual page (would require WebFetch, not granted to this lane), and the author's claimed command outputs (markdownlint, detect.test.sh, changelog-parity scripts) — those are author-claimed and unverified, not confirmed by me.

This is a well-scoped, internally consistent documentation update that closes a previously-recorded fetch gap without touching detector logic. No blocking or suggestion-level issues found.

[claude]

Claude finished @kyle-sexton's task in 33s —— View job

---

Security Review — skipped (no security-relevant surface)

• Gather context (PR diff, changed files)
• Apply skip gate check
• Review diff for security-relevant surface
• Post findings (or clean bill of health)

This PR touches only three files, all documentation/metadata:

• plugins/ai-slop/.claude-plugin/plugin.json — version bump 0.3.1 → 0.3.2.
• plugins/ai-slop/CHANGELOG.md — new changelog entry (prose).
• plugins/ai-slop/skills/audit/reference/catalog.md — reference catalog prose closing a documented "fetch gap" with quoted Wikipedia content and citations.

No code, scripts, GitHub Actions workflows, config parsing logic, or dependency pins are modified. There's no executable surface, no user input handling, no auth/access-control logic, and no injection-relevant code paths in this diff. This meets the security-review skip gate ("change has no security-relevant surface after reading the diff") — no findings to report.

[github-actions]

Last security-reviewed head: 5abbec36bd66e1a998af6ac3be3c04084e8c4f9a. On the next push, the relevance gate compares only the commits since this SHA; delete this comment to force a full re-review.

[github-actions]

Claude has reviewed this PR 3 times. The lane skips further automatic reviews after 5; deleting this comment resets the count.

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 5abbec36bd

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]

Claude finished @cursor[bot]'s task in 2m 11s —— View job

---

Code review (/review:code-review)

• Gather context (PR diff at HEAD 094a870, description, changed files)
• Run /review:code-review per custom instructions
• Post findings as inline comments (if any)
• Post summary comment

Scope

Reviewed the full diff 42cf1fd...094a870 (3 files, +151/-12 — matches the PR's declared stats). This head adds a second commit on top of the one already reviewed by this lane (5abbec3): docs(ai-slop): slug the Comment-specific indicators as recorded-only.

Findings

None that a careful reviewer would block or flag on the current head.

The new commit is a direct, correct fix for the one open review comment on this PR (Codex, P2: "Add catalog entries for the comment-specific indicators"). Verified independently:

• All 6 previously-untracked Comment-specific tells now have proper ### rule-<slug>: <name> entries (rule-misquoted-policies, rule-maintenance-banner-transclusion, rule-sectioned-comments, rule-request-critic-input, rule-dismiss-origin-speculation, rule-redirect-to-content), each with detectability / applicability / v1 fields matching the file's existing convention (compared against neighboring entries e.g. rule-unused-named-refs). The 7th tell correctly stays a cross-reference to the already-slugged rule-canned-policy-assurance rather than a duplicate entry.
• Inventory count math checks out mechanically: grep -c '^### rule-' on the full file returns 72, matching "65 Wikipedia-sourced + 7 Cursor unslop additions" stated in the Inventory section.
• grep -c '^- v1: script$' returns 15, matching the "15 shipped script rules" claim used in both the Ineffective-indicators verdict and the CHANGELOG's "no script-roster change" claim.
• No new slug collides with an existing one (checked the full ### rule- list); rule-sectioned-comments is distinct in scope from the pre-existing rule-section-summaries (AfC draft summaries vs. talk-page comment sectioning) and from rule-verbose-edit-summaries, which the new entry explicitly cross-references.
• CHANGELOG.md's 0.3.2 entry was updated in lockstep with the catalog change (six slugs now named explicitly, inventory delta 59→65 stated).

Not independently re-verified (no tool access in this lane / sandbox blocked network use for this run): the exact wording of the live/pinned Wikipedia quotes against the actual page, and the author's claimed command outputs (markdownlint-cli2, detect.test.sh, changelog-parity scripts) — those remain author-claimed and unverified by me.

[claude]

Claude finished @cursor[bot]'s task in 1m 35s —— View job

---

Code review (/review:code-review)

• Gather context (PR diff origin/main...HEAD, description, changed files)
• Run /review:code-review per custom instructions
• Post findings as inline comments (if any)
• Post summary comment

Scope

Reviewed the full diff at HEAD f341148 (3 files, +151/-12): plugins/ai-slop/skills/audit/reference/catalog.md, plugins/ai-slop/CHANGELOG.md, plugins/ai-slop/.claude-plugin/plugin.json. Docs/metadata only — no detector, emitter, eval, or workflow code touched, so REVIEW.md's code-review-lane checks (audit-log-on-security-action, multi-write atomicity) don't apply here — nothing security-relevant or multi-write in this change.

Findings

None that block or need a fix. Independently verified the load-bearing claims rather than taking them on faith:

• 72 ### rule- entries total (grep -c '^### rule-') = 65 + 7, matching the Inventory line.
• 15 - v1: script slugs in the catalog, and independently 15 distinct rule-* matches in detect.sh — corroborates the "no shipped rule appears here" verdict against the Ineffective-indicators list from an independent angle (the roster's own source file), not just the catalog's self-count.
• rule-em-dash sits under ## Style (L221–280) and rule-rule-of-three under ## Language and grammar (L170–219) — matches the claim that both remain valid signs on other sections of the source page, not Ineffective indicators.
• Comment-specific indicators: 1 pre-existing (rule-canned-policy-assurance, already under ## Edit summaries) + 6 new recorded-only/wikipedia-specific entries = 7, matching the section's own count and closing the P2 Codex flagged on the prior commit (094a870).
• CHANGELOG [0.3.2] heading format and prose match the [0.3.1] entry's convention; plugin.json version (0.3.2) matches.
• The em-dash quote's own em dash ("in LLM detection—and may even indicate the opposite") is inside a blockquote in a file covered by the `` marker at the top of catalog.md, so it won't trip the detector on itself.

Not independently re-run (no network/build tool access in this sandbox lane): markdownlint-cli2, detect.test.sh, and the changelog-parity scripts the PR description cites — those remain author-claimed. Also not re-verified: the exact wording of the live Wikipedia quotes against the actual page (no WebFetch in this lane).

This is a well-scoped, internally consistent fetch-gap closure. No inline comments posted — the diff has nothing to flag at the line level.