Issues

• Release tests: the fake forge's environments are sometimes unreadable under load

  bug
  cego/collie#201 · Filter by author madskronborg opened on Oct 23, 2026

• A parked release cannot resume after its branch moved past the head it last saw

  bug
  cego/collie#200 · Filter by author madskronborg opened on Oct 23, 2026
  1

• codex-stop: 'a continuation that started before the server took its drain' fails under CI load with extra interrupts

  bug
  cego/collie#199 · Filter by author madskronborg opened on Oct 23, 2026

• An implement Run that does not start from a bare plan directory tells its reviewer there is no spec, though its own plan/SPEC.md exists

  bug
  cego/collie#196 · Filter by author madskronborg opened on Oct 23, 2026

• Configure Claude subagent models, cache retention, and delegation per managed agent

  ready-for-agent
  cego/collie#193 · Filter by author madskronborg opened on Oct 23, 2026

• A review offers "Fix findings" from a checkout that refuses every fix, and asks again

  cego/collie#185 · Filter by author madskronborg opened on Oct 22, 2026

• The criteria loop hands back a build that reported a blocker and changed nothing, then fails the Run instead of asking

  cego/collie#183 · Filter by author madskronborg opened on Oct 22, 2026

• A stopped Run carried on after a host restart and filed issues, with no resume

  cego/collie#176 · Filter by author madskronborg opened on Oct 21, 2026
  1

• A gate fix that changed nothing and disputed every gap does not end the fix loop

  cego/collie#175 · Filter by author madskronborg opened on Oct 21, 2026

• An implement Run that cites its own plan Run's files as evidence fails outside this Run

  cego/collie#174 · Filter by author madskronborg opened on Oct 21, 2026

• Every self-improve Improvement fails to start: its plan child is refused "then" is not an input

  cego/collie#173 · Filter by author madskronborg opened on Oct 21, 2026

• A plan's wrong outcome kind holds its implement Run to a proof the work cannot have, and nothing can correct it

  cego/collie#172 · Filter by author madskronborg opened on Oct 21, 2026

• Semantic judgements on a branch past the evidence cap are paid for at every boundary and can only answer unverified

  cego/collie#167 · Filter by author madskronborg opened on Oct 21, 2026
  1

• A Run with no checkout always ends aligned: unverified, because its goal judgement needs a git diff

  cego/collie#166 · Filter by author madskronborg opened on Oct 21, 2026
  2

• Plan tickets require a command to pass without knowing it already fails on the base branch

  cego/collie#165 · Filter by author madskronborg opened on Oct 21, 2026
  1

• A Run reports aligned: true — every constraint was checked when nothing was judged

  cego/collie#164 · Filter by author madskronborg opened on Oct 21, 2026
  1

• Cross-run judgement reads every sibling's commit in the judging Run's own repository, so a multi-repository wave never sees its siblings' work

  cego/collie#163 · Filter by author madskronborg opened on Oct 21, 2026
  2

• A steer sent to a just-launched agent before its step makes the step fail with an empty reason

  cego/collie#162 · Filter by author madskronborg opened on Oct 21, 2026

• A Run whose agent never got its work fails for good instead of parking for resume

  cego/collie#161 · Filter by author madskronborg opened on Oct 21, 2026
  1

• A Run whose agent died with its Machine stays running, and resume does not start the agent again

  cego/collie#160 · Filter by author madskronborg opened on Oct 21, 2026

• An implement Run started from another Run's plan directory gets a fresh branch named plan, not that Run's branch

  cego/collie#159 · Filter by author madskronborg opened on Oct 21, 2026