Skip to content
Back to skills

Quick Refactor

ASecurity

Cleanup pass on a change once its PR is open, before the Codex review. Triggers: 'quick refactor', 'quick-refactor', 'リファクタリング', 'refactor what we just did'. Forked Opus subagent at high effort.

  • 3 stars
  • 0 votes
  • 0 copies
  • 0 views
  • Added September 21, 2026
ai-agentsgosqlgitfrontendbackend

Security analysis

A100/100

Scanned September 21, 2026

npx -y skills add aitit-inc/leadace --skill quick-refactor --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Quick Refactor?

Add the live security badge to your README. It updates with every re-scan.

Security grade badge for Quick Refactor
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/aitit-inc-quick-refactor/badge)](https://www.skillsdirectory.com/skills/aitit-inc-quick-refactor)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

Download with Pro
SKILL.md
---
name: quick-refactor
description: "Cleanup pass on a change once its PR is open, before the Codex review. Triggers: 'quick refactor', 'quick-refactor', 'リファクタリング', 'refactor what we just did'. Forked Opus subagent at high effort."
context: fork
model: opus
effort: high
---

# Quick refactor

Cleanup of a change whose PR is open, before the Codex review. Behaviour does not change.

## Scope

- Baseline: uncommitted changes (staged, unstaged, untracked). None → `git diff develop...HEAD`.
- `$ARGUMENTS`: file names narrow the baseline; a commit range or PR number replaces it.
- Read each hunk with enough context to know what it is for.

## Pass

Changed content only, in this order:

1. **Comments** — delete by default. Keep one only for a *why* the code cannot carry, or where dropping it would invite the bug it prevents. As short as it can be.
2. **Waste** — remove what the change does not need. Aim for the smallest, most obvious implementation that meets the spec. Fewer lines is not the goal; no abstraction for its own sake.
3. **Symptomatic fixes** — a guard that patches one case → write the rule instead. Only while the edit stays inside the change; anything wider goes in the report.
4. **LLM-facing text** — prompts, skill / reference text, tool descriptions: shorten without changing meaning. No filler; plain words, short sentences.

Untouched: content the change did not introduce (mention it in the report).

## Verify and report

- Run the area's checks from CLAUDE.md (backend: `npm run typecheck`, `npm test`; frontend: `npm run check`). Changed SQL predicate → its `e2e/regression-*.sh` if the local stack is up.
- Do not commit, push, or merge.
- Report: what was removed or reshaped and why; what was left as out of scope; check output verbatim on failure.

Attribution

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

Comments

Loading comments…

Newsletter

Get the weekly list of safe new skills, and the ones we pulled.