Skip to content

feat(release): require exact canary evidence - #40

Merged
vriesd merged 4 commits into
mainfrom
codex/eval-phase-9
Aug 25, 2026
Merged

feat(release): require exact canary evidence#40
vriesd merged 4 commits into
mainfrom
codex/eval-phase-9

Conversation

@vriesd

@vriesd vriesd commented Aug 25, 2026

Copy link
Copy Markdown
Contributor

Summary

  • add exact-artifact OpenCode canary prepare, record, and verify commands
  • persist strict passed, failed, or incomplete records with checklist, expiry, ActorIdentity, and sanitized evidence bindings
  • include canarySha256 in final decisionInputSha256 and keep canary-bound decisions immutable and distinct
  • verify readiness on main with read-only permissions and publish only tags after strict exact decision plus canary checks

Verification

  • bun run check: 520 pass, 1 skip, 0 fail
  • package smoke: 1 pass
  • actionlint: green
  • exact fixture auto-load verified on OpenCode 1.18.6 with flow-reviewer, flow-auto, and flow-status present
  • dry-run: INCONCLUSIVE for missing canary, exit 0
  • strict mode: missing canary blocks, exit nonzero
  • four-model Interrogate: no unresolved infrastructure blockers

Manual stop gate

The maintainer-run OpenCode canary and resulting canary-bound VERIFIED decision are intentionally not fabricated. Phase 9 release readiness remains INCONCLUSIVE until those reviewed files are committed. No tag or release is requested.

@vriesd
vriesd merged commit 889e80e into main Aug 25, 2026
6 checks passed

@chatgpt-codex-connector chatgpt-codex-connector Bot left a comment

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 0702650f18

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".

Comment thread scripts/eval-canary.ts
"canary",
`${prepared.artifact.packageVersion}.json`,
);
await writeImmutable(path, Buffer.from(canonicalJson(record)));

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

P1 Badge Keep canary evidence out of the artifact source identity

When the documented prepare → record → qualify flow is followed, this writes the canary into the repository after prepareCanary captured its ArtifactIdentity; the session/transcript files are written there as well. The subsequent qualification rebuild calls inspectArtifact, whose source digest includes tracked and untracked repository files, so it rejects the canary as belonging to a different artifact, and committing these records also changes the sourceCommit checked by the tag workflow. Consequently no canary produced by this command can satisfy the new strict release gate; the artifact identity must remain reproducible after evidence is persisted.

Useful? React with 👍 / 👎.

Comment thread scripts/eval-canary.ts
Comment on lines +389 to +393
return mapStrings(normalized, (text) =>
scrubSecrets(text).replace(
/\b(?:ses_[A-Za-z0-9]+|(?:session|review):[A-Za-z0-9-]+)\b/g,
"<redacted-id>",
),

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

P1 Badge Redact values associated with sensitive field names

When exported session or transcript JSON contains an opaque credential under a sensitive key, such as { "password": "correct-horse-battery-staple" }, this maps and scrubs each key and value independently, so scrubSecrets never sees the key/value association and the value does not match any token-prefix pattern. The plaintext is then written to the committed canary artifact despite being described as sanitized; use field-aware recursive redaction in addition to string-pattern scrubbing.

Useful? React with 👍 / 👎.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants