| name | writing-plans |
| description | Use after the design is clear or approved and before non-trivial implementation. Break the work into small executable slices with exact files, validations, and exit criteria. |
Writing Plans
Use this skill to convert a clear design into a plan that can be executed without guessing.
Planning Rules
- Break work into small, coherent slices.
- Name the exact files, symbols, or tests when known.
- Give each slice a concrete validation command.
- Call out risks to repository invariants or user-visible behavior.
- Prefer the next smallest useful step over broad batches.
Helper
Use the scaffold when it saves time:
bash .claude/skills/superpowerz/scripts/new_plan.sh tmp/superpowerz/specs/YYYY-MM-DD-feature-slug.md
Default location:
tmp/superpowerz/plans/YYYY-MM-DD-feature-slug.md
Minimum Plan Shape
Every plan should answer:
- what slice is first
- what file or symbol controls that slice
- what check can falsify the current hypothesis
- what makes the slice done
Execution Handoff
Once the plan exists, execute one slice at a time.
After the first substantive edit, the next action should be focused validation when one exists.
Use .claude/skills/executing-plans/SKILL.md as the default execution workflow instead of ad hoc task hopping.
Exit Condition
The plan is complete when the next implementation step is obvious, local, testable, and ready to be carried out by .claude/skills/executing-plans/SKILL.md.
