/.github/loop/PRIORITIES.md

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)

1. Expose fallback_echo during A2A streaming fallback conformance (#3560) — PR #3731 closed the plan/delegate completion regression, so this is now the top remaining Now-phase interop/conformance gap. It protects the A2A streaming promise developers see in the README and site by ensuring the non-native streaming fallback path receives the tool surface, while staying scoped and CI-verifiable.
2. Parse multi-event A2A SSE fallback responses in the harness (#3662) — once the fallback tool path succeeds, the harness must accept legitimate multi-event message/stream responses instead of concatenating valid SSE events into invalid JSON. This is a small CI-verifiable harness fix that keeps cross-provider streaming conformance focused on real gateway failures rather than parser brittleness.
3. Propagate agent run cancellation and deadlines through model and tool calls (#3544) — the highest-value remaining resilience gap is predictable failure semantics across agent runs, model calls, tool calls, plan/delegate, and flow handoffs. Tool retries, live-provider deadline tuning, and delegated-plan completion are in place; the lifecycle still needs cancellation/deadline propagation so work fails safely instead of becoming opaque loops.
4. Emit OpenTelemetry spans for agent run timelines (#3525) — recent work made runs inspectable, correlated trace metadata through scheduled dispatch, verified restart resume, added opt-in tool retries, hardened provider conformance, and fixed provider-emitted text tool calls. The next Next-phase step is to turn that RunInfo foundation into standard OTel spans for agent runs, model calls, tool calls, checkpoint/resume, cancellation/deadlines, and failures.
5. Add an AP2 mandate layer over A2A and x402 (#3552) — this is a forward interop investment, not a Now-phase blocker: Go Micro already has A2A agents and x402 paid tools, so a small signed-mandate foundation can keep agent payments aligned with the open-protocol story without pulling the queue away from adoption, resilience, or observability. Keep it additive and opt-in while the AP2/FIDO work settles.

Seeded by Claude Code from the roadmap + open issues; thereafter maintained by the architecture-review pass.