Repository navigation
docs(upstream): re-audit addyosmani agent-skills drift and move the pin to 39b723c - #16526
Conversation
…in to 39b723c Re-audits docs/upstream/addyosmani-agent-skills.md from pin 301b02b to upstream head 39b723c (41 commits, 2026-07-10 to 2026-11-14). Every row link moves to the new SHA; four evals/README.md anchors move with the six lines inserted above them (L43 to L49, L63 to L69, L65 to L71, L71 to L77). Rows the drift run named re-read unchanged: the cited sections of code-review-and-quality, constraint-driven-development, debugging, planning, security-and-hardening, test-driven-development, the security checklist, /build, the eval runner and the link validator kept their meaning; their dispositions stand. Eight new rows record the ideas the diff adds: doc staleness from commit order and append-only change logs, project-tooling verification and APG patterns for UI, a version label with a deferred list, delta specs for a shipped feature, the /build source-spec guard (all already have), and the OWASP Top 10:2025 renumbering (adapt, not applied: the security-reviewer floor still uses 2021 ids, and an A10:2025 row waits on a recorded miss). The pasted-content adapt row and the safe-fallback reject row gain the new evidence. The license notice is re-read at the new pin: MIT, unchanged. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Codex Review Summary: ✅ CompletedThis comment shows the latest Codex review activity on this pull request.
ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings. |
Lane run report
Failed GitHub API calls
|
|
claude-review reviewed Code review (incremental, since
|
|
claude-security-review reviewed Security review: PR #16526 (incremental, blocking only)No security issues found. This was an incremental review covering only the change since What I readI read all of What changedOne cell changed in the row on OWASP Top 10:2025 numbering. The old text said the 2021 list's "A04, A06 and A10 ids name other categories". The new text says "A02 to A06 and A10 ids name other categories". The rest of the row is unchanged, and its disposition is still Security checks
This change is documentation only and has no security-relevant surface (skip gate 2). Reading the diff confirms it removes no control. critical-findings: 0 |
The OWASP 2025 row said only A04, A06 and A10 name other categories. Against the published 2025 list, A02 through A06 and A10 do. Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
|
Reply to the claude-review summary on
The fix names A02 to A06 and A10 in the row and in the body. Not taken from the finding: the statement that Vulnerable Components is folded into A03:2025, which I did not verify against the category page. |
|
Finding 1 is fixed in 67e1092: the OWASP 2025 row now names A02 to A06 and A10, and the body's Remaining work says the same. |
Refs: #16488
Summary
Re-audits
docs/upstream/addyosmani-agent-skills.mdfor drift issue #16403. The page moves from pin301b02bto upstream head39b723c(2026-11-14). That span holds 41 commits (compare)./buildspec selection, the eval runner's resource staging, a slug fix in the reference-link validator, and the manifest versions.already have: staleness from commit order, append-only change logs, verifying UI with the project's own tooling and the APG patterns, a version label with a deferred list, delta specs for a shipped feature, and the/buildsource-spec guard. One isadapt, not yet applied: OWASP Top 10:2025 numbering.adaptrow adds the "answer looks obvious from the pasted text" probes. The safe-fallbackrejectrow notes that upstream's own session sample now fails closed.needs-humanissue was filed.needs-human, and the drift run's sweep closes it once the moved pin reads clean.Fix
301b02b12bff98e4688e6984af6556802322bbdeto39b723ceebf81474d0557c1fecd01eddab585563.evals/README.mdline 43:security-and-hardening#L138changed only its OWASP id.documentation-and-adrsrows, onefrontend-ui-engineering, one OWASP 2025, twospec-driven-development, one/build auto.Disposition counts over the 46 rows #16403 names, plus the 7 new rows (53 in all):
None of the 46 named rows changes its disposition.
Remaining
On #16488 after this pull request (it delivers the addyosmani/agent-skills re-audit, drift issue #16403):
On this page's own adapt row:
plugins/review/agents/security-reviewer.md:62-75(the OWASP Top 10 floor) keeps the 2021 ids and does not name the edition. Under the 2025 list, A02 to A06 and A10 name other categories, and SSRF sits inside A01. Moving the table to 2025 ids changes no check. A row for A10:2025 (Mishandling of Exceptional Conditions) is a new check class, so under the page's eval-first decision it waits for a recorded miss. It is not applied here, because doing so would pullplugins/reviewinto a docs-only re-audit.Verification
scripts/check-upstream-drift.sh --page docs/upstream/addyosmani-agent-skills.md:upstream records: 1 pages, no driftscripts/check-upstream-drift.sh --links --page docs/upstream/addyosmani-agent-skills.md:upstream records: 243 links ok across 1 pagesscripts/check-cost-figures.sh --check: no cost figure outside the baselined and exempt filesscripts/check-affected.sh: 13 passed, 5 skipped over 1 changed filegh api repos/addyosmani/agent-skills/compare/301b02b...39b723c, file by file, for every path Upstream drift: addyosmani/agent-skills #16403 lists.39b723cis still the newest commit on upstreammain.file:linecitation into this repository was read atorigin/main8cdfe3e.check-line-citations.pyskipsdocs/upstream/records by design, so those citations were checked by hand.Related
🤖 Generated with Claude Code