fix(governance): let the hypatia baseline filter actually run (scan exit-on-findings under bash -e)#464
Merged
Conversation
…indings) Follow-up to #455. That PR wired apply-baseline.sh into validate-hypatia-baseline but missed that `hypatia-cli.sh scan .` exits non-zero *by design* whenever it finds anything — and the step runs under `bash -e`, so errexit aborted the step at the scan line, before the relativization + apply-baseline.sh (the real gate) ever ran. (This is also why the pre-#455 job was always red: its dead `jq` count never executed either.) Caught by verifying the fix end-to-end on neurophone#171, where the job died at "scan complete: N findings; exit 1" with no filter output. Fix: tolerate the scan's own exit code (`|| true`) so the filter runs and gates instead — but guard against a genuine scanner crash being swallowed into a false pass by requiring the scan output to be a valid JSON array first (fail loudly if not; no silent-green). Both behaviours simulated locally under `bash -e`. Note (separate, repo-side): a newer hypatia HEAD now reports more findings than when a given repo's .hypatia-baseline.json was written, so a caller may still see its gate stay red on genuinely-new findings after this lands — that is the filter working correctly, and is resolved per-repo by extending the baseline (or fixing at source), not here. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_0172RBMz3qYjb1ttzD2i7RNh
|
hyperpolymath
added a commit
that referenced
this pull request
Jul 3, 2026
…checkout tooling) (#466) ## Summary Third and final piece of the `validate-hypatia-baseline` fix (after #455 wired the filter and #464 made it run). The gate checks out standards' `scripts/` into `.standards-checkout/` to get `apply-baseline.sh` — but that directory sits **inside the caller's working tree**, so `hypatia-cli.sh scan .` scanned it too and reported **standards' own files as the caller's findings**. Caught on neurophone#172, where 4 of the 8 "unfiltered" findings were pure tooling pollution: ``` .standards-checkout/scripts/check-ts-allowlist.ts banned_language (critical) .standards-checkout/neurosym-a2ml/scripts/bootstrap.sh shell_download (high) .standards-checkout/agentic-a2ml/scripts/bootstrap.sh shell_download (high) .standards-checkout/.../rsr-deployer/scripts/bootstrap.sh shell_download (high) ``` None is the caller's code. **Every** adopting repo would inherit these. ## The fix ```bash cp .standards-checkout/scripts/apply-baseline.sh "$RUNNER_TEMP/apply-baseline.sh" rm -rf .standards-checkout # <- before the scan HYPATIA_FORMAT=json … scan . … # now only sees the caller's own tree … bash "$RUNNER_TEMP/apply-baseline.sh" … ``` Copy the filter out to `$RUNNER_TEMP` (outside the workspace), delete `.standards-checkout/`, then scan. hypatia only ever sees the repo under audit; the tooling can't pollute the finding set. ## Net effect With **#455** (baseline actually consulted) + **#464** (filter actually runs) + **this** (clean scan tree), the estate gate finally behaves as intended: suppress a repo's acknowledged baseline entries, fail on genuinely-new findings, and never conflate the tooling with the repo. Opt-in per repo via SHA-pin as before. Draft pending CI + your review. (As with #455/#464, this job **skips** on standards' own CI — no baseline here — so it was reasoned through against neurophone#172's real output.) 🤖 Generated with [Claude Code](https://claude.com/claude-code) https://claude.ai/code/session_0172RBMz3qYjb1ttzD2i7RNh --- _Generated by [Claude Code](https://claude.ai/code/session_0172RBMz3qYjb1ttzD2i7RNh)_ Co-authored-by: Claude <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Follow-up to #455. That PR wired
apply-baseline.shintovalidate-hypatia-baseline, but the fix never actually ran:hypatia-cli.sh scan .exits non-zero by design whenever it finds anything, and the step runs underbash -e, so errexit aborted the step at the scan line — before the relativization +apply-baseline.sh(the real gate) could run.This is also why the pre-#455 job was always red on every repo with findings: its dead
jqcount never executed either. The scan's exit-1 killed the step every time.Caught by verifying end-to-end on neurophone#171, where the job died at:
…with no
apply-baseline.shoutput at all.The fix
|| true— tolerate the scan's by-design exit-on-findings so the baseline filter runs and becomes the gate.Both behaviours simulated locally under
bash -e:Heads-up for callers (separate, repo-side)
A newer hypatia HEAD reports more findings than when a repo's
.hypatia-baseline.jsonwas written (neurophone: 18 now vs 14 when its baseline was authored in #168). So after this lands, a caller may still see its gate stay red on genuinely-new findings — that is the filter working correctly. Resolution is per-repo: extend the baseline (or fix at source), not here.Draft pending CI + your review (edits the shared security gate). Same opt-in-via-SHA-pin safety as #455.
🤖 Generated with Claude Code
https://claude.ai/code/session_0172RBMz3qYjb1ttzD2i7RNh
Generated by Claude Code