Prepare for a sprint planning discussion when no active sprint exists.
Steps:
Read context:
- Read
docs/BACKLOG.mdfor all candidate work items. - Read
MEMORY.mdfor recent changes, architecture state, and user preferences. - Check
git log --oneline -10for recent commits. - Scan
docs/sprints/complete/to understand the arc of recent sprints (what themes have been in focus, what was deferred).
- Read
Categorize backlog items into groups such as:
- Infrastructure / plumbing — internal improvements, build system, CI, container infra, credential management.
- Agent behavior / instructions — how agents receive and interpret tasks, concern handling, role-specific issues.
- Orchestrator features — scheduling, retry, gating, multi-graph, observability.
- Extensibility / future-proofing — new frameworks, new environments, Rust migration, multi-model support.
- Documentation / testing gaps — missing docs, e2e coverage holes.
- Use your judgment on categories — these are suggestions, not fixed.
Suggest a priority ordering with rationale:
- Which items build on recently shipped work (momentum)?
- Which items unblock other items (dependencies)?
- Which items reduce risk or friction for ongoing e2e testing?
- Which items are quick wins vs. multi-PR efforts?
- Flag any items that seem stale or already addressed.
Present to the user:
- A categorized summary of the backlog (not the full text — concise descriptions with section references).
- A suggested "top 3–5 items" for the next sprint, with reasoning.
- Any items you'd recommend removing or merging with others.
- Ask the user which direction they want to go before drafting a sprint doc.