docs(claude): replace the required-context count with the measured list - #1965
docs(claude): replace the required-context count with the measured list#1965seonghobae wants to merge 1 commit into
Conversation
CLAUDE.md said `.github` uses "classic branch protection with 14 named
required contexts". The live endpoint returns 12.
gh api repos/ContextualWisdomLab/.github/branches/main/protection \
--jq '.required_status_checks.contexts | length'
12
Measured 2026-09-06: CodeQL compatibility analysis (actions), CodeQL
compatibility analysis (python), Detect CodeQL languages, coverage-evidence,
dependency-review, noema-review, opencode-review, osv-scan,
required-workflow-bootstrap, scan-pr-queue, scorecard, trivy-fs.
Not "14 became 12". `git log -S "named required contexts" -- CLAUDE.md`
returns exactly one commit, 7d09388 (#1766), and the sentence has never been
revised since. Branch-protection edits leave no trace in git and the REST API
exposes no history, so whether the count was wrong when written or two
contexts were later removed is not recoverable. The text says so instead of
guessing, so a later reader does not infer a removal that has no evidence.
Records the list rather than a bare number: a count cannot be checked against
anything, while a list can be diffed against the endpoint and shows *which*
context drifted. The bare number went stale unnoticed for three days.
The count also cannot be corroborated from inside the repository. CHANGELOG
line 52 already says "0/12 required contexts", but that figure came from
7f4c5e3 (#1937) — the same author's own census — so it is the same
measurement read twice, not an independent source. The protection endpoint is
the only authority here, which is exactly why the text tells the reader to
re-read it.
Evidence:
- uvx ruff check --select F821 scripts/ci tests: All checks passed
- full suite: 2956 passed, 0 failed
- coverage: TOTAL 100%; interrogate: PASSED (minimum 100.0%)
- no test pins this prose: `git grep '14 named\|named required contexts'`
over tests/ and scripts/ returns nothing
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
|
Warning Review limit reachedNext included review available in 40 minutes. View limit detailsLimit details: You’ve used the included review currently available. You've used all free OSS reviews for now. Wait for the free limit to reset to keep reviewing this public repository. Review configuration: ⚙️ Run configurationConfiguration used: Organization UI Review profile: CHILL Plan: Team Run ID: 📒 Files selected for processing (1)
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
Codex Review SummaryThis 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. |
![devin-ai-integration[bot]](https://avatars.githubusercontent.com/in/811515?s=60&v=4){kind=small}
검증 근거 (교차 세션)peer 세션(
방법론에 대한 한 가지 명시peer 와 저자가 각각 protection 엔드포인트를 조회한 것은 독립 측정 두 건이 아닙니다 — 같은 아티팩트를 두 번 읽은 것입니다. 이 사실에는 그보다 권위 있는 소스가 존재하지 않으므로 본문에서도 "corroborated" 가 아니라 "the only authority" 로 표현했고, 그래서 문서가 독자에게 목록을 믿지 말고 엔드포인트를 다시 읽으라고 지시합니다. 저자(이 세션)와 검증(peer 세션)은 의도적으로 분리했습니다. 🤖 Generated with Claude Code |
1 participant
CLAUDE.mdsaid.githubuses "classic branch protection with 14 named required contexts". The live endpoint returns 12.Measured 2026-09-06:
CodeQL compatibility analysis (actions),CodeQL compatibility analysis (python),Detect CodeQL languages,coverage-evidence,dependency-review,noema-review,opencode-review,osv-scan,required-workflow-bootstrap,scan-pr-queue,scorecard,trivy-fs.This is not "14 became 12"
git log -S "named required contexts" -- CLAUDE.mdreturns exactly one commit —7d093881(#1766) — and the sentence has never been revised since. Branch-protection edits leave no trace in git and the REST API exposes no history, so whether the count was wrong when written or two contexts were later removed is not recoverable. The new text says that explicitly rather than guessing, so a later reader does not infer a removal that has no evidence behind it.Why a list instead of a corrected number
A bare count cannot be checked against anything, so it rots silently — this one did, for three days, in the file every agent reads first. A list can be diffed against the endpoint, and it shows which context drifted rather than only that something did. The paragraph now also tells the reader to re-read the endpoint rather than trust the list, because a stale list here cannot be detected by reading the repository.
On corroboration — deliberately not claimed
CHANGELOG.mdline 52 already says "0/12 required contexts", which looks like independent agreement. It is not. That figure came from7f4c5e3e(#1937), the same author's own census, so treating it as a second source would be one measurement read twice. The protection endpoint is the only authority for this fact — which is exactly why the text points the reader back at it.Evidence
uvx ruff check --select F821 scripts/ci tests— All checks passedcoverageTOTAL 100%;interrogatePASSED (minimum 100.0%)git grep "14 named\|named required contexts"overtests/andscripts/returns nothingReviewer note
Found while adjudicating four workflow-concurrency questions for a peer session; the peer independently re-measured the endpoint and got the same 12, and confirmed the
git log -Sresult. Authoring and verification are deliberately split across sessions — the peer is attached on the verification side.Documentation only. No workflow, script, or test behaviour changes.
🤖 Generated with Claude Code