I pointed my code reviewer at its own verifier. It found two ways to lie.
说明与用途
SeamStress 发现并修复了两个关键的验证门缺陷,确保只有真正有证据支持的发现才会被标记为已验证。
问题
SeamStress, a code reviewer, incorrectly marked findings as verified without proper evidence and sometimes attached proof to the wrong finding due to namespace collisions.
结果
After fixing the verification gate, SeamStress now reliably reports only findings backed by real evidence and prevents proof misattribution, enhancing trust in its audit process.
工作流
- Run SeamStress on its own verification gate using the standard pipeline (three blind critics, synthesis, per‑finding verification).
- Identify defects: (1) verified status without evidence, (2) proof attached to wrong finding due to slug collision.
- Apply fixes: (1) add evidence presence check in the authority logic; (2) key namespace on seam position to avoid slug collisions.
- Run regression tests (src/types/types.test.ts, src/engine/detector.test.ts) to confirm fixes.
- Publish updated code and documentation.
- Maintain open‑source repo and benchmark data for transparency.
证据
The status authority looked like this: const result = verifications.find((v)=> v.findingId === finding.id); return result ? result.status : "unverified";
The fix lives at the authority, not just the schema: if (!result) return "unverified"; const hasRealEvidence = result.evidence.some((e)=> e.quotedCode.trim().length > 0); return hasRealEvidence ? result.status : "unverified";
- 记录日期: 2026-09-10T00:30:16.182Z