メニュー
Use when a written plan exists and implementation should proceed in small validated slices. Carry out one plan slice at a time, keep scope tight, validate immediately after edits, and do not drift away from the approved plan without calling it out.
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
SOC 職業分類に基づく
Create new skills, modify and improve existing skills, and measure skill performance. Use when users want to create a skill from scratch, update or optimize an existing skill, run evals to test a skill, benchmark skill performance with variance analysis, or optimize a skill's description for better triggering accuracy.
Local fallback entrypoint for the Superpowers plugin. Use when the user mentions Superpowers, the plugin is unavailable, or you need to route non-trivial engineering work in this repository into the local brainstorming, planning, debugging, and verification skills.
Use before implementing new behavior, refactors with meaningful design choices, or changes that alter user-visible behavior. Clarify intent, propose approaches, write a short spec, and only then transition to planning.
Use for bugs, regressions, failing tests, broken scripts, or mismatches between expected and actual behavior. Start from a concrete anchor, form one falsifiable local hypothesis, make a small grounded fix, and validate immediately.
Use when starting non-trivial engineering work in this repository, when the user mentions Superpowers, or when the plugin is unavailable and you need to route the task into the local skill suite before acting.
Use when about to claim something is fixed, complete, passing, reviewed, or ready. Require fresh command output and report the observed result rather than the intended result.
| name | executing-plans |
| description | Use when a written plan exists and implementation should proceed in small validated slices. Carry out one plan slice at a time, keep scope tight, validate immediately after edits, and do not drift away from the approved plan without calling it out. |
Use this skill after a design is clear and a plan exists.
This skill governs execution discipline. It is for carrying out the plan without wandering, batching unrelated edits, or claiming progress without evidence.
Prefer this order when more than one check is available:
Track slices in the todo tool while executing.
At any given time, exactly one slice should be in progress.
While executing, preserve repository invariants and keep canonical history separate from derived state.
If a planned change would violate replayability, provenance, append-only integrity, or governance traceability, stop and revise the plan before editing.
Exit this skill only after the active slice has validation evidence and the next state is clear:
.claude/skills/verification-before-completion/SKILL.md for a completion claim