fix(education): correct changelog entries and restore teach action hint

[kyle-sexton]

No related issue: audit finding fix

Summary

Audit finding plugin-education (changelog-integrity, low): the education 0.11.6 changelog entry claimed more than the change did, the 0.11.3 entry ended in a bare (#4119) instead of a link, and the teach argument hint had dropped its action list. Ships as education 0.11.7.

Fix

• plugins/education/CHANGELOG.md: [0.11.6] now says only that the examples moved into the skill body (the old hint carried only (e.g., ...) examples); [0.11.3] links #4119 like 0.11.4 to 0.11.6; new [0.11.7] entry.
• plugins/education/skills/teach/SKILL.md: argument-hint lists the closed action set from the Action Router table (95 characters, inside the 100-character budget).
• plugins/education/.claude-plugin/plugin.json: version 0.11.6 to 0.11.7.
• plugins/education/skills/quiz-me/evals/evals.json case 9 lists a new auth-middleware diff fixture (evals/fixtures/auth-middleware-change/auth-middleware.diff: public-path guard, rate-limit-before-auth ordering, invalid-token 401 path) and names those behaviors in its first expectation, so the sandbox has a change to quiz on (planning request, Eval candidates: behavioral heuristics from the Finding Your Unknowns integration #3589 AC2).
• plugins/education/skills/setup/evals/evals.json case 3: the escaped em dash in the prompt is replaced by a colon (work-items request, F10 of unit 4070).
• In-place corrections to released entries: [0.11.6] reworded to say only that the examples moved into the skill body; [0.11.3] issue reference linked. Both are repeated in the [0.11.7] entry, as scripts/check-changelog-parity.sh requires.

Verification

• bash scripts/check-changelog-parity.sh --check --check-order: pass.
• bash scripts/check-changelog-parity.sh --check-bump origin/main: pass.
• bash scripts/check-changelog-parity.sh --check-preserved origin/main: pass (45 headings compared).
• bash scripts/validate-plugins.sh: all manifests and the catalog validated.
• bash scripts/validate-plugin-contracts.test.sh: PASS=92 FAIL=0, zero argument-hint warnings in the shipping tree.
• education has no test scripts of its own.

Related

• Audit report: .work/audit/REPORT.md, finding plugin-education.
• argument-hint hygiene: six concrete fixes, a house style rule, and a lint criterion #3542 (argument-hint house style, delivered earlier), I32 routes to nowhere: 14 skill-body references name skills this marketplace does not ship #4119 (linked in the 0.11.3 entry).
• Cross-group requests applied: work-items F10 (unit 4070, test(evals): three-case floor for fifteen thin skills (#4070) #4838) and planning quiz-me case 9 (Eval candidates: behavioral heuristics from the Finding Your Unknowns integration #3589 AC2). A paid with/without eval run of case 9 stays with the owner, so Eval candidates: behavioral heuristics from the Finding Your Unknowns integration #3589 is Refs only.
• Verification of both: the two evals.json files parse, plugins/skill-quality/scripts/check-evals-quality.sh passes on each, and the changelog parity checks pass.

🤖 Generated with Claude Code

https://claude.ai/code/session_01EugXnFddtpHcY5gTuyEirB

[chatgpt-codex-connector]

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review	Status	Commit	Review trigger
📝 Code Review	✅ Completed 2026-09-29T13:26:30.224028Z	467b7ab	Draft marked ready

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" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.

[claude]

Claude finished @kyle-sexton's task in 31s —— View job

---

Security review complete ✅

• Gather context (PR diff, changed files)
• Run /review:security-review criteria against the diff
• Report findings

No security issues found.

This PR is documentation/metadata only: a CHANGELOG.md correction, a plugin.json version bump (0.11.6 → 0.11.7), an argument-hint string change in teach/SKILL.md, and eval fixture/prompt edits in quiz-me and setup. None of it touches authentication, authorization, secrets handling, or user-input processing in the plugin's own logic.

Two points I specifically checked and ruled out:

• New fixture evals/fixtures/auth-middleware-change/auth-middleware.diff (permalink) is inert quiz content read by the quiz-me eval sandbox for question-generation, not code that gets executed or interpreted by the plugin.
• Instruction-surface lens: the argument-hint and CHANGELOG.md edits don't remove or soften any standing guardrail (no rule, allowlist, or hook block-list is touched), so the instruction-exception-register check doesn't apply here.

[claude]

Claude finished @kyle-sexton's task in 1m 41s —— View job

---

Claude finished the code-review lane

• Gather context (PR body, changed files)
• Invoke /review:code-review and follow its procedure
• Post inline findings (if any) via inline comments
• Post final summary

Summary

Reviewed all 6 changed files against origin/main...HEAD (467b7ab). This is a low-risk changelog/eval/argument-hint correction with no meaningful bugs found:

• plugins/education/skills/teach/SKILL.md — new argument-hint (<topic|codebase|mission|glossary|resources|explain|primer|exercise|assess|resume|status> [args]) is exactly 95 characters (confirmed by counting) and lists all 11 actions in the Action Router table (lines 97-109) with no omissions or extras.
• plugins/education/CHANGELOG.md — [0.11.7] entry is well-formed, dated, and links #3589; [0.11.6] and [0.11.3] in-place corrections read correctly and are consistent with the repeated corrections in [0.11.7] (matches the check-changelog-parity.sh --check-preserved requirement described in the PR body).
• plugins/education/.claude-plugin/plugin.json — version bump to 0.11.7 matches the new changelog heading.
• plugins/education/skills/quiz-me/evals/evals.json case 9 + new fixture evals/fixtures/auth-middleware-change/auth-middleware.diff — the fixture diff genuinely contains all three behaviors the updated expectation names: a PUBLIC_PATHS guard, rateLimit middleware moved before auth (ordering), and a caught/converted invalid_token 401 instead of a thrown error. The files path convention (evals/fixtures/...) matches existing usage in plugins/bugs/skills/write/evals/evals.json and plugins/overengineering/skills/justify/evals/evals.json.
• plugins/education/skills/setup/evals/evals.json case 3 — em dash replaced with a colon, no other change.

No high-signal correctness or maintainability issues found on the changed lines, so no inline comments were posted.

Note on verification: this sandbox's permission mode denied bash scripts/check-changelog-parity.sh ..., validate-plugins.sh, and even a plain python3 -c "json.load(...)" sanity check, so I could not independently re-run the checks the PR body cites. I did visually confirm both evals.json diffs are structurally well-formed (matching brackets/commas, no dangling entries) via direct file reads. The PR's own script-run results are author-claimed and unverified by me.

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 467b7ab20c

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

---

Code review lane — skipped (already reviewed)

• Gather context (PR body, changed files, existing review comments)
• Invoke /review:code-review and follow its procedure
• Check skip gate
• Post final summary

Per the /review:code-review skip gate, I stopped before a full re-review: this exact head, 467b7ab20c2b08a72050f6a227b12858e75182bc, already has a successful pass from this lane — the "Claude finished the code-review lane" comment from earlier today reviewed all 6 changed files against origin/main...HEAD (467b7ab) and found no high-signal correctness/maintainability issues. Since the head hasn't moved since that pass, it still applies, so I'm not duplicating it with a second lane pass or new inline comments.

For completeness, note that chatgpt-codex-connector (a separate automated reviewer, not this lane) flagged two items on this same head that this lane's criteria would also consider in-scope as correctness bugs in the eval fixtures:

• plugins/education/skills/setup/evals/evals.json:31 — the prompt's trailing colon (/education:setup: set quiz_policy...) makes the slash-command token parse as /education:setup: rather than /education:setup, so the case could exercise command-routing failure instead of the intended refusal behavior.
• plugins/education/skills/quiz-me/evals/fixtures/auth-middleware-change/auth-middleware.diff:5 — the first hunk header's line counts (-1,24 +1,41) don't match the actual hunk body (13 old / 32 new lines), which git apply --numstat rejects as a corrupt patch (corrupt patch at line 48), so diff-tooling-based eval runs can't parse the fixture as intended.

These are genuine, verifiable defects (I confirmed the header/line-count mismatch and the trailing-colon parse by reading the files directly) worth fixing before merge, but I'm surfacing rather than re-reporting them as this lane's own findings since they're already on the PR from Codex and this lane's pass on this head is complete.

[claude]

Claude finished @kyle-sexton's task in 40s —— View job

---

Security review complete ✅

• Gather context (PR diff, changed files, custom instructions)
• Run /review:security-review criteria against the diff at HEAD 467b7ab
• Report findings

No security issues found.

This PR touches 6 files, all metadata, documentation, and eval-fixture content for the education plugin — no runtime logic, auth, authorization, secret handling, or user-input processing is changed:

• plugins/education/.claude-plugin/plugin.json — version bump only.
• plugins/education/CHANGELOG.md — text corrections and a new [0.11.7] entry.
• plugins/education/skills/teach/SKILL.md — argument-hint string restored to the closed action list (permalink). No guardrail, allowlist, or hook block-list is touched, so the instruction-surface-deletion lens doesn't apply.
• plugins/education/skills/quiz-me/evals/evals.json + new evals/fixtures/auth-middleware-change/auth-middleware.diff (permalink) — the fixture is fictional quiz content (src/middleware/auth.js) consumed only as text by the quiz-me eval sandbox for question generation. It is never parsed as a real diff, applied, executed, or interpreted by plugin logic, so the malformed hunk header Codex flagged (line-count mismatch) and the security-shaped pattern inside the fixture text (rate-limit-before-auth ordering) carry no exploit path in this repo.
• plugins/education/skills/setup/evals/evals.json — prompt punctuation change only (em dash → colon); no code path.

No CRITICAL / IMPORTANT / SUGGESTION findings.

[claude]

Claude finished @kyle-sexton's task in 1m 21s —— View job

---

Code review complete ✅

• Gather context (PR body, changed files, existing comments)
• Invoke /review:code-review and follow its procedure
• Post inline findings via inline comments
• Post final summary

No findings.

This lane's prior pass (on 467b7ab) is stale — the head moved to aaff99a, so I re-reviewed the full diff (origin/main...HEAD, 6 files) fresh rather than relying on the earlier "no findings" result.

Commit aaff99a fixes both defects Codex flagged on 467b7ab, and I independently confirmed both fixes by inspection:

• evals/fixtures/auth-middleware-change/auth-middleware.diff — the first hunk header now reads @@ -1,13 +1,34 @@ (was -1,24 +1,41). I hand-counted the hunk body: 5 context + 8 removed = 13 old lines, 5 context + 29 added = 34 new lines, matching the header exactly. Blank context lines also got their leading space restored. The second hunk (@@ -9,7 +9,7 @@) checks out the same way (6 context + 1 removed = 7 old, 6 context + 1 added = 7 new).
• plugins/education/skills/setup/evals/evals.json:31 — the prompt is now /education:setup set quiz_policy... (no trailing colon), so the slash-command token parses as /education:setup cleanly.

I also re-checked the rest of the diff for anything new:

• teach/SKILL.md argument-hint: exactly 95 characters (counted), and its 11 listed actions match the Action Router table (SKILL.md#L99-L109) exactly — no extras, no omissions.
• CHANGELOG.md: version headings remain in descending order (0.11.7 → 0.11.2), the [0.11.6] and [0.11.3] headings themselves are unchanged (only their body text was reworded/linked), and the new [0.11.7] entry restates those same corrections as the parity script requires.
• plugin.json: version bump (0.11.6 → 0.11.7) matches the new changelog heading.
• quiz-me/evals/evals.json case 9: files path (evals/fixtures/auth-middleware-change/auth-middleware.diff) matches the convention used elsewhere (e.g. plugins/bugs/skills/write/evals/evals.json), and the expectation text names exactly the three behaviors the fixture now contains (public-path guard, rate-limit-before-auth ordering, invalid-token → 401).

Not independently verified: I don't have shell permission in this session to run scripts/check-changelog-parity.sh or git apply against the fixture, so the PR body's cited script runs remain author-claimed. I instead hand-verified the hunk arithmetic and changelog heading order directly from file contents, which is sufficient to confirm the two previously-flagged defects are fixed and no new issues were introduced.
· branch fix/audit-education

[claude]

Claude finished @kyle-sexton's task in 1m 4s —— View job

---

Security review complete ✅

• Gather context (PR body, changed files, custom instructions)
• Invoke /review:security-review criteria against the diff at HEAD aaff99a
• Report findings

No security issues found.

This head (aaff99a) adds one commit on top of the previously-reviewed 467b7ab: it fixes the two non-security defects Codex flagged in that earlier review pass. I verified both fixes directly:

• auth-middleware.diff — the first hunk header now reads @@ -1,13 +1,34 @@. I recounted the hunk body by hand: 13 old-file lines (3 context + 1 added-context + 8 removed + 2 trailing context — wait, precisely: lines 6,8,9 context, 10‑17 removed, 46‑47 context = 13) and 34 new-file lines (context + the 28 added lines + trailing context), both matching the header. The stray unprefixed blank context lines are also gone. This is inert eval-fixture text consumed only as quiz content by the quiz-me sandbox, never applied or executed by plugin logic — so the earlier malformed-patch issue was a fixture-quality bug, not a security bug, and it's now fixed.
• setup/evals/evals.json:31 — the prompt is now /education:setup set quiz_policy... with a space, not a colon, after the command token. No command-routing ambiguity remains.

No other files changed since the prior review at 467b7ab, which already covered the plugin.json version bump, the CHANGELOG.md corrections, and the teach/SKILL.md argument-hint restoration — none of that touches auth, secrets, or user-input processing in plugin logic, and nothing in this PR narrows or removes a standing instruction/guardrail (the instruction-surface-deletion lens doesn't apply).

[claude]

Claude finished @kyle-sexton's task in 45s —— View job

---

Security review complete ✅

• Gather context (PR diff, changed files, custom instructions)
• Check skip gate (prior review staleness vs HEAD 16618ac)
• Run /review:security-review criteria against the diff
• Report findings

No security issues found.

Reviewed the full diff (origin/main...HEAD, HEAD 16618ac2a0a05d070f8985f18c2742dfbda74dbf), which is one commit past the last-reviewed head (aaff99a). That final commit (16618ac) only strips the leading-space blank context lines from the auth-middleware.diff eval fixture (an editorconfig trailing-whitespace fix) and re-tightens the hunk headers to match — no new files, no logic, no security surface.

All 6 changed files remain metadata/documentation/eval-fixture content for the education plugin, with no runtime logic, auth, authorization, secret handling, or user-input processing:

• plugins/education/.claude-plugin/plugin.json — version bump only (0.11.6 → 0.11.7).
• plugins/education/CHANGELOG.md — text corrections and a new [0.11.7] entry.
• plugins/education/skills/teach/SKILL.md#L4 — argument-hint restored to the closed action list. This narrows/clarifies user-facing guidance; it doesn't remove or soften a guardrail, allowlist, or hook block-list, so the instruction-surface-deletion lens doesn't apply.
• plugins/education/skills/quiz-me/evals/evals.json + new fixture auth-middleware.diff — the fixture is fictional quiz content (src/middleware/auth.js) read only as text by the quiz-me eval sandbox for question generation. It's never applied, executed, or interpreted by plugin logic, so neither the security-shaped pattern inside the fixture text (rate-limit-before-auth ordering, token revocation) nor its now-fixed hunk-header arithmetic carries an exploit path in this repo.
• plugins/education/skills/setup/evals/evals.json#L31 — prompt punctuation change only (colon → space after the slash-command token); no code path.

No CRITICAL / IMPORTANT / SUGGESTION findings.
· branch fix/audit-education

[claude]

Claude finished @kyle-sexton's task in 1m 44s —— View job

---

Code review complete ✅

• Gather context (PR body, changed files, existing comments)
• Invoke /review:code-review and follow its procedure
• Post inline findings (if any) via inline comments
• Post final summary

No findings.

The prior pass on this lane (aaff99a) is stale — head moved to 16618ac — so I re-reviewed the full diff (gh pr diff 5243, 6 files) fresh.

The only change since aaff99a is commit 16618ac ("drop blank context lines from the quiz-me diff fixture"), a cosmetic-only edit to auth-middleware.diff: added lines that are blank now carry no trailing space (avoids the trim_trailing_whitespace = true rule at .editorconfig:17), with no change to the tested behavior. I hand-verified the current fixture is a well-formed unified diff:

• Hunk 1 header @@ -1,11 +1,32 @@: counted 11 old lines (3 context + 8 removed) and 32 new lines (3 context + 29 added) — matches.
• Hunk 2 header @@ -9,5 +9,5 @@: counted 5 old / 5 new — matches.
• No blank context line lacks its leading space (checked with cat -A).

This resolves both defects Codex flagged on 467b7ab (malformed hunk headers, trailing colon in the setup eval prompt) — both were already fixed in aaff99a and remain fixed here.

I also re-checked the rest of the diff:

• teach/SKILL.md argument-hint: counted 95 characters, inside the 100-char budget. Its 11 listed actions (topic|codebase|mission|glossary|resources|explain|primer|exercise|assess|resume|status) match the Action Router table rows exactly, same order, no omissions/extras.
• quiz-me/evals/evals.json case 9: files references evals/fixtures/auth-middleware-change/auth-middleware.diff, and the fixture genuinely contains all three behaviors the updated expectation names — the PUBLIC_PATHS guard, rateLimit moved before auth (ordering), and a caught invalid_token → 401 instead of a thrown error.
• setup/evals/evals.json case 3: prompt is now /education:setup set quiz_policy to always in settings.json for me. — no trailing colon, parses as a clean /education:setup command token.
• CHANGELOG.md / plugin.json: [0.11.7] heading, corrected [0.11.6]/[0.11.3] bodies, and the version bump are internally consistent.

Not independently verified: shell access in this session denied git fetch, git apply --check, and even python3 -c "json.load(...)", so I could not execute the PR body's cited check-changelog-parity.sh / validate-plugins.sh runs or a git apply dry-run against the fixture. Those remain author-claimed. I instead hand-verified the hunk arithmetic, JSON structure (via gh pr diff output — matching brackets/commas, no dangling entries), and character counts directly, which is sufficient to confirm no regressions from aaff99a and no new issues in 16618ac.
· branch fix/audit-education

[claude]

Claude finished @kyle-sexton's task in 59s —— View job

---

Security review complete ✅

• Gather context (PR diff, changed files, custom instructions)
• Check skip gate (prior review staleness vs HEAD 9404d81)
• Run /review:security-review criteria against the diff
• Report findings

No security issues found.

Skip gate applies. HEAD 9404d81c097858631a00d36b3d6cbbfdcdf77cf5 is a merge of origin/main into this branch on top of the already-reviewed 16618ac. git diff 16618ac 9404d81 --stat shows 119 files touched, all of them unrelated plugins pulled in from main (e.g. plugins/animation/, plugins/biome-format/, plugins/discovery/, plugins/typos-format/) — none under plugins/education/. gh pr diff 5243 confirms the PR's actual base-relative diff is unchanged: still the same 6 files as the prior full review at 16618ac (job run), which found no security issues:

• plugins/education/.claude-plugin/plugin.json — version bump only.
• plugins/education/CHANGELOG.md — text corrections and a new [0.11.7] entry.
• plugins/education/skills/teach/SKILL.md#L4 — argument-hint restored to the closed action list. No guardrail, allowlist, or hook block-list is touched, so the instruction-surface-deletion lens doesn't apply.
• plugins/education/skills/quiz-me/evals/evals.json + new fixture auth-middleware.diff — fictional quiz content read only as text by the quiz-me eval sandbox for question generation; never applied, executed, or interpreted by plugin logic. No exploit path.
• plugins/education/skills/setup/evals/evals.json#L31 — prompt punctuation change only; no code path.

Codex's two prior findings on this PR (malformed diff-fixture hunk headers, a trailing colon in the setup eval prompt) are eval-fixture correctness issues, not security, and were already fixed in commits aaff99a/16618ac.

No CRITICAL / IMPORTANT / SUGGESTION findings.
· branch fix/audit-education

[claude]

Claude finished @kyle-sexton's task in 37s —— View job

---

Code review — skipped (already reviewed)

• Gather context (PR body, changed files, existing comments)
• Invoke /review:code-review and follow its procedure
• Check skip gate
• Post final summary

Per the /review:code-review skip gate, I stopped before a full re-review. Current HEAD is 9404d81c097858631a00d36b3d6cbbfdcdf77cf5, a merge of origin/main into this branch. I diffed the plugins/education/ tree between the last-reviewed head (16618ac, reviewed in the "Code review complete" comment above — no findings) and 9404d81: it is empty. The merge only pulled in unrelated main commits (dfc6cad, 55e6e9b, 638ecd3, 2c871ec, …); none of the 6 files this PR touches changed.

That prior pass already confirmed:

• plugins/education/skills/teach/SKILL.md — argument-hint is 95 chars, matches the Action Router table's 11 actions exactly.
• plugins/education/CHANGELOG.md / plugin.json — [0.11.7] heading, corrected [0.11.6]/[0.11.3] bodies, and the version bump are internally consistent.
• plugins/education/skills/quiz-me/evals/evals.json + evals/fixtures/auth-middleware-change/auth-middleware.diff — fixture is a well-formed unified diff (hunk headers @@ -1,11 +1,32 @@ and @@ -9,5 +9,5 @@ both verified against hunk bodies) containing all three behaviors the updated case-9 expectation names.
• plugins/education/skills/setup/evals/evals.json — case 3 prompt has no trailing colon after the command token.

Both defects Codex flagged on the original head (467b7ab) — the malformed diff-fixture hunk headers and the trailing-colon prompt — were fixed in aaff99a and remain fixed. No new findings on this head.
· branch fix/audit-education