chore(engine): subtree pull opencode 08ddbdc (m29 scorecard guard) - #44
Conversation
08ddbdc test(dfir): offline scorecard regression guard (fixture-driven) (#19) d5641ec feat(opencode): emit used_fallback from native LLM runtime gate (#20) 136158a test(dfir): offline scorecard regression guard (fixture-driven) git-subtree-dir: engine git-subtree-split: 08ddbdc668a4cfddfd35efc74e247a5bc7f381c5
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: d5f4cd3be6
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".
| } | ||
| }, | ||
| "attack-samples": { | ||
| "expected_verdict_band": "INDETERMINATE", |
There was a problem hiding this comment.
For the attack-samples case, this runtime-side oracle disagrees with the authoritative scorecard in fixtures/dfir-scorecard/ground-truth.json, and the existing fixtures/dfir-scorecard/runs/attack-samples.verdict.json is actually SUSPICIOUS. If this case is added to the new Bun scorecard guard or reused by callers, the grader will reject the known-good sealed run for the wrong reason instead of catching real regressions.
Useful? React with 👍 / 👎.
1 participant
Summary
mainat08ddbdc(test(smoke): clarify agent hang timeout after /v1 fix #19 offline scorecard regression guard) intoengine/— squash commita5deded, merged5f4cd3.verdictbinary: sha256a9197243e48a39c9c3a495fc316655f4fd63d69e572dbdf410250cee744d50d2;used_fallbackevidenced in built runtime (native-runtime test 1 pass/0 fail; 12 string hits in binary).Test plan
npm run typecheckordered (SDK → TUI → CLI): ok (lane logtmp/m29-logs/engine.log)node scripts/selftest.mjson this branch: 216 passed, 0 failed (orchestrator re-run after fullnpm run build)Honest residual
/home/...lines remain in engine custody fixtures (paths under/home/raven/caseforge-core/evidence/...); fix is dev-repo-owned path relativization, out of scope here.m29 lane: engine (grok headless), worktree
cf-m29-engine, baseorigin/mainfeef841.