Codex
2b2a3fda12
2.6 KiB
Priorities
The ranked work queue for the autonomous improvement loop. The architecture-review pass (the architect) owns this file: each run it turns the roadmap plus an internal scan (gaps in the services → agents → workflows lifecycle, API coherence, drift, tech debt, test and DX friction) into a single ordered list — highest-value first — and links each item to a tracking issue. The hourly continuous-improvement pass works the top item whose issue is still open. So the architect decides what, and the increment loop builds it.
Reading / editing. An item is done when its linked issue closes (the increment
that builds it adds Closes #<issue>). Roadmap phase (Now → Next → Later) is the
primary ordering; internal findings are interleaved by value, not kept in a
separate list. The human can reorder this list — or the issues — at any time to
redirect the loop; direction always wins.
Off-limits to the loop (the architect proposes these as notes, never as queue items the loop can auto-merge): brand/positioning copy, breaking public-API changes, architectural rewrites. Those go to the human.
Work queue (ranked)
- Add CI-verified first-agent chat/inspect fixture (#4644) — With plan/delegate persistence hardened and the first-agent breadcrumbs refreshed, the highest-value remaining adoption gap is proving the real CLI-shaped scaffold → run → chat → inspect loop for the smallest provider-free agent. Keep this focused and separate from the richer support 0→hero reference so a new developer can trust the first-agent path without provider keys.
- Add scheduled cross-provider agent conformance (#4649) — The Now roadmap still calls for battle-testing the same agent scenario across providers. After the no-secret first-agent fixture, verify tool-calling, multi-step turns, plan/delegate, and guardrails on the live-provider matrix behind key-gated scheduled CI, without slowing the local harness.
- Harden agent provider failure resilience (#4650) — The next reliability seam is making timeouts, cancellation, rate limits, retry/backoff, and inspectable failure metadata predictable through the agent loop. Keep this scoped to existing behavior and tests; surface any breaking API or default changes as human review notes instead of queue work.
Seeded by Claude Code from the roadmap + open issues; thereafter maintained by the architecture-review pass.