Use this skill only when creating or improving an ExecPlan for work with real sequencing constraints, migration risk, or nontrivial validation requirements.
Do not use this skill for:
- simple bug fixes
- small isolated edits
- straightforward tasks with no real sequencing constraints
Goal
Improve the quality of the final ExecPlan by doing a compact decomposition first.
The decomposition is a means, not the output. The output should be a better ExecPlan in this repository's expected format.
Required workflow
- Read
AGENTS.md,ARCHITECTURE.md, anddocs/PLANS.mdfirst. - Carry forward:
- hard constraints
- major risks and unknowns
- sequencing implications
- validation requirements
- Identify the task structure:
- prerequisite or foundation work
- branches that can proceed independently
- integration or synthesis points
- validation, observability, rollout, and rollback work
- Convert that structure into the ExecPlan:
- milestones in dependency order
- explicit assumptions and unknowns
- concrete verification commands
- recovery or rollback notes when relevant
- Prefer repository-specific grounding over abstract planning language.
- If uncertainty remains, encode it as an investigation milestone rather than pretending it is resolved.