Commit 6022e49
fix(registry): regenerate to clear accumulated drift (registry-verify red on main) (#366)
## What
Regenerates `.machine_readable/REGISTRY.a2ml` so the `registry-verify`
gate is green on `main` again. Purely recorded `source_hash` values —
**no source or generator changes**.
## Why `main` is red
Two merges changed content **under spec homes** without regenerating the
registry, so the recorded hashes no longer match the tree
(`build-registry.sh --check` fails — reproduced on `bec6161`):
| Merge | What it changed | Specs whose `source_hash` drifted |
|---|---|---|
| #356 | edited `hypatia-rules/registry-staleness.a2ml` |
`hypatia-rules` |
| #361 | added `timeout-minutes` to in-home `.github/workflows/*` |
`a2ml`, `k9-svc`, `meta-a2ml`, `0-ai-gatekeeper-protocol`,
`avow-protocol`, `axel-protocol`, `consent-aware-http`,
`rhodium-standard-repositories` |
(The `source_hash` is `sha256(git ls-files -s <home>)`, so **any**
tracked-file change under a home — including a vendored
`.github/workflows/*.yml` — changes it.)
## The fix
`just registry` (`scripts/build-registry.sh`) → refreshes the **9**
affected hashes. Rebased onto current `main` (`bec6161`); supersedes the
earlier single-hash version of this PR.
## Verification
- Before: `build-registry.sh --check` → **DRIFT** (exit 1) on `main`.
- After: **OK** (exit 0).
## Systemic note (not fixed here)
This keeps recurring because PRs that touch a spec home are being merged
without running `just registry`, and `registry-verify` evidently isn't
blocking the merge. Worth making `registry-verify` a required check (or
adding a pre-merge regenerate) so the registry can't drift on `main`
again. Happy to open that as a follow-up if you want.
https://claude.ai/code/session_0178nMYCNXgotTeekePkoUjd
Co-authored-by: Claude <noreply@anthropic.com>1 parent c97b2bd commit 6022e49
0 file changed
0 commit comments