wassnameandClaudypoo 16827de45d judge goes strictly read-only: no bash, never re-runs verify; reviews evidence discipline instead ...

Re-running verify was fine for 'npm test' but a footgun for ML workflows where verify may be a
10-hour training run -- and bash made 'read-only' nominal anyway (it could mutate). The agent now
runs verify itself and saves the output as evidence. The judge checks, in order: anything here /
quoted+attributed / provenance / quotes match disk / substance. Matches the cooperative-but-
confused threat model: reading real artifacts catches confusion; execution only defended against
deliberate forgery, which is out of scope.

Co-Authored-By: Claudypoo <288921227+claudypoo@users.noreply.github.com>

2026-07-03 12:13:14 +08:00

..

append-log.test.ts

v2: delete the parser -- plan.md is for LLMs, the judge subsumes the machinery

2026-07-03 10:24:35 +08:00

decide-signoff.test.ts

judge: extract decideSignOff so inconclusive always means 'ran but failed'

2026-07-03 10:55:45 +08:00

judge-args.test.ts

judge goes strictly read-only: no bash, never re-runs verify; reviews evidence discipline instead

2026-07-03 12:13:14 +08:00

tick-goal.test.ts

sign-off ticks the goal [x] itself + document that the judge reads the working tree, not HEAD

2026-07-03 11:49:31 +08:00