- name
- decompose-goals
- description
- Use two modes to understand existing work or decompose proposed tasks into criteria, specifications, and decisions. For autonomous analysis and planning; use decompose-with-user for collaborative shaping or decompose-with-agents for an actual agent exchange.
# Decompose goals autonomously
Read [the shared guidance](references/principles.md), then the mode that fits the request:
- **Understand existing work:** [mode guidance](references/modes/understand.md).
- **Decompose proposed work into criteria:** [mode guidance](references/modes/decompose.md).
Use both when needed, keeping findings about the present distinct from proposals for change. Each mode links the relevant templates. Keep a decision/interview log by default; coding work produces specifications and ADRs.
Investigate consequential questions you can answer yourself. Use reasoned defaults for reversible choices within your authority and make material assumptions visible. A small experiment can be better than an elaborate plan.
Challenge your understanding with concrete counterexamples and evidence. Self-review is useful but is not independent verification. Do not invent stakeholder answers. A decision that belongs to the user may need their input without stopping independent authorized work.
Keep a persistent decision log of the questions and choices that shape the work. In understanding mode, record what the evidence establishes. In decomposition mode, develop criteria and a specification; coding work includes ADRs as described in the mode guidance. Use relevant available skills and repository templates when they improve the result.
Return the requested result, its supporting records, and meaningful uncertainty. Planning does not itself authorize execution or starting a persistent goal.
在 GitHub 查看