| name | writing-plans |
| version | 6.2.0 |
| category | workflow |
| description | Use when an approved change spans multiple dependent outcomes and needs a durable implementation sequence before coding. Do NOT use for small, obvious, reversible edits. |
Writing Plans
Write plans that preserve decisions and dependencies without expanding every
edit into a ceremony.
Read ../using-superpowers/references/risk-policy.md.
When a Plan Is Worth It
Create a plan for multi-component work, migrations, public contracts, risky
rollouts, or work likely to continue across sessions. Skip it for a
straightforward change that fits in a short active checklist.
Plan Shape
Save durable plans to the project convention, otherwise
docs/superpowers/plans/YYYY-MM-DD-<topic>.md.
Include:
- Goal and non-goals.
- Relevant existing boundaries and exact files.
- Tasks grouped by independently useful outcome.
- Dependencies and interfaces between tasks.
- Risk tier for each task.
- The smallest verification that proves each outcome.
- Rollback or recovery for high-risk changes.
Do not include placeholder steps, invented APIs, repeated prose, routine shell
commands, or a commit after every tiny action. Include code only where an exact
contract or non-obvious algorithm must be preserved.
Test and Review Budget
- Low risk: existing check or smoke path; no new test by default.
- Medium risk: one focused behavior test only when it protects a key outcome.
- High risk: minimal test-first regression or contract coverage.
- Independent review appears once at the completed-change boundary only when
risk or the user requires it.
Execution Handoff
Inline execution is the default. Use superpowers:subagent-driven-development
only when delegation was requested or tasks are genuinely independent.