This repo uses Dots for cross-session planning. The tracked state lives in .dots/, is committed with the repo, and uses the opencode prefix.
Start each session with:
dot ls
dot readyBefore starting work:
dot on <id>When the work is done:
dot off <id> -r "What changed"- Use top-level epics only to group related work. Pick child dots for actual execution.
- Use
-adependencies when work must wait on another dot. - Keep upstream sync, PR intake, replay, testing, release gating, and infrastructure work as separate dots so the audit trail stays clear.
- Prefer one end-to-end deliverable per dot. Split follow-up work into child dots instead of letting one dot sprawl.
- Close every completed dot with a concrete reason.
0: incidents, security, broken release pipeline1: core PR triage and integration automation2: release gates, supporting infra, and tooling3: docs and cleanup4: parking lot and umbrella epics
The initial backlog is seeded in .dots/:
Launch autonomous PR integrationas the umbrella epicDefine PR rubricIngest upstream PR metadataReplay candidate PRs on devGate fork promotionsUpgrade Actions runtimes
Replay candidate PRs on dev is intentionally blocked on the rubric and metadata tasks so the first implementation pass has clear inputs.