ワンクリックで
slipway-review
Inspect intent and quality once and report findings without modifying code or creating a repair loop.
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。
メニュー
Inspect intent and quality once and report findings without modifying code or creating a repair loop.
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
Resolve genuine human decisions one question at a time after investigating facts.
Explicitly split an Objective into self-contained vertical Changes with native relationships.
Apply authorized code changes and report implementation-time technical activities honestly.
Explicitly materialize clarified requirements as one self-contained Change or one planning Objective.
Explicitly start and drive Slipway's interruptible soft autopilot one structured Action at a time.
| name | slipway-review |
| description | Inspect intent and quality once and report findings without modifying code or creating a repair loop. |
| disable-model-invocation | true |
Use this capability only for a current Review Action or when the user explicitly invokes it directly. It is always read-only.
For a Run Action, compare the current workspace with the immutable Run-start HEAD and fingerprint. List pre-existing dirty and untracked paths separately. A start-to-current difference is only an observation: concurrent user edits, other Runs, or tools may have contributed, so never claim that the current Run or Implement caused it merely because it appeared later.
Inspect two dimensions:
Report concrete findings with locations, details, and uncertainties. The strict Run Outcome must carry action_kind: "review" matching the current Action. Submit only these combinations:
completed with review.result no_findings_reported or findings_reported;partial with review.result inconclusive;error with review.result error.Set pause and implementation to JSON null, include every common Outcome field, and leave suggested_actions empty. not_run is reserved for a CLI review-skip projection.
Do not modify files, return needs_input, ask the user to choose a repair, suggest Implement, issue pass/fail or approval language, pause the Run, or create an automatic repair or re-review loop. Findings always flow to Summary; the user may later choose a separate Change or a new Run.
When invoked directly outside a Run, use the baseline named by the user. If none is named, inspect the current workspace against HEAD and state that it is a standalone baseline, not a Run-start snapshot. Return the same factual review without inventing Run attribution or protocol IDs.