SikongJueluo 0637784f5f feat(ai-judge): conversation evidence with bounded whitelist capture ...

- add conversation.ts: compaction-aware active-branch capture, user-text-only whitelist, 16-item and 12,000-char bounds with latest-user preservation
- bump prompt to bash-shadow-v2 with explicit-user-intent authority rules and quoted untrusted intent evidence
- capture the requesting cwd and per-ask conversation state; flip evidence-quality flags from placeholders to measured values
- record the candidate-identity change for prior cohorts in the scenario-set doc

2026-08-17 22:25:06 +08:00

..

analyzer

docs(research): archive shadow replay rounds and analyzer round-1 fixes

2026-08-17 18:59:19 +08:00

config.test.ts

feat(ai-judge): global config module with validation and cohort identity

2026-08-17 19:53:48 +08:00

conversation.test.ts

feat(ai-judge): conversation evidence with bounded whitelist capture

2026-08-17 22:25:06 +08:00

evidence.test.ts

refactor(permission): consume structured bash payload

2026-08-16 23:08:45 +08:00

judge.test.ts

feat(ai-judge): review sink with telemetry health and fail-closed truth table

2026-08-17 21:37:48 +08:00

lifecycle.test.ts

feat(ai-judge): conversation evidence with bounded whitelist capture

2026-08-17 22:25:06 +08:00

model.test.ts

fix(ai-judge): classify provider aborts after timeout as timeout

2026-08-17 16:58:10 +08:00

prompt.test.ts

refactor(permission): consume structured bash payload

2026-08-16 23:08:45 +08:00

review.test.ts

feat(ai-judge): review sink with telemetry health and fail-closed truth table

2026-08-17 21:37:48 +08:00