メニュー
Reflect on session learnings and update a skill so future agents succeed without prior context. Use when the user asks to update or refine a skill, or when this session surfaced dead ends, correct pathways, or coverage gaps worth capturing.
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。
Full PR review workflows -- first-pass, deeper (overcoverage), follow-up (re-review). Load when the user invokes /review-pr or explicitly asks for a PR review, deeper wave, or re-review; never ambiently.
Shared atom for analyzing a diff and writing a structured PR description. Background knowledge for workflow commands -- not invoked directly.
Author a publication-grade single-file HTML design doc (RFCs, architecture proposals, sync prep). Editorial typography, decision cards, fact-checked code references, Playwright-verified layout, secret-gist publishing. Use when an engineer asks for a design doc, RFC, monograph, sync-prep brief, or "make this look like a real spec."
Shared atom for running quality checks, committing, and pushing. Background knowledge for workflow commands -- not invoked directly.
Self-review your aggregate diff before shipping to catch dead weight, junk code, perf misses, pattern drift, AI slop, and scope creep. Use when the user asks to scrub, stranger-review, or clean up changes prior to opening a PR.
Shared atom for structural maintainability judgment on a diff -- hunt code-judo simplifications, spaghetti-condition growth, file-size explosions, and orchestration smells. Background knowledge for review and self-review flows -- not invoked directly.
SOC 職業分類に基づく
| name | update-skill |
| description | Reflect on session learnings and update a skill so future agents succeed without prior context. Use when the user asks to update or refine a skill, or when this session surfaced dead ends, correct pathways, or coverage gaps worth capturing. |
| argument-hint | <skill-name> [context about what was learned] |
Review the work done in this session (or as described in $ARGUMENTS). Identify:
Rewrite/amend the skill so that a future agent with no prior context would succeed:
Round 1: Read the updated skill as a fresh agent. Is it clear? Complete? Would you succeed following only these instructions? Fix any issues.
Round 2: Check for redundancy, ambiguity, missing edge cases, inconsistent formatting, and ordering. Fix any issues.
Final Round: Critique and refine holistically, x2, including optimizing for information density per token so we are not bloating the content/size(s) needlessly, till you are happy with and proud of the work you've done <3
