メニュー
Visual QA, screenshot review, browser review, Playwright, responsive QA, accessibility QA, UI polish, design review, visual regression, anti-AI UI rubric. Use when validating frontend visuals or reviewing screenshots.
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
| name | visual-qa-rubric |
| description | Visual QA, screenshot review, browser review, Playwright, responsive QA, accessibility QA, UI polish, design review, visual regression, anti-AI UI rubric. Use when validating frontend visuals or reviewing screenshots. |
Gunakan untuk mengecek apakah UI layak ship secara visual, aksesibilitas, responsive, dan tidak terlihat generated AI.
Jika browser tool tersedia:
Jika browser tool tidak tersedia:
Skor tiap kategori:
Kategori wajib:
Skor interpretasi:
Jika “ya” untuk 2+ item, UI perlu product-specific polish.
## Visual QA Report
**Surface:** ...
**Evidence:** screenshot/browser/static review
| Category | Score | Finding | Fix |
|----------|-------|---------|-----|
**Total:** x/20
**Verdict:** ship-ready | polish-needed | risky | redesign-needed
## Browser Evidence
- Desktop:
- Tablet:
- Mobile:
- Console/network:
## Required Fixes
1. ...
Framework-agnostic website security audit and authorized pentest methodology (OWASP, red-team attacker mindset, white-hat reporting). Use for web pentest, vulnerability assessment, attack-surface review, bug bounty prep, or security audit of any web stack.
Critical path identification, risk-adjusted effort sizing, complexity detection, and execution strategy selection. Use when scoping a task, deciding direct-vs-planned execution, breaking down large work, or assessing how many steps something really needs.
Multi-phase workflows, state machines, DAG task execution, checkpoint/resume, and saga patterns for complex multi-step or multi-session work. Use when a task spans 5+ steps, multiple files, multiple sessions, or has irreversible phases.
Understanding legacy and unfamiliar code, git blame/history analysis, dependency archaeology, and reconstructing intent behind existing code. Use when working in unfamiliar codebases, investigating "why was this written this way", or before changing code you didn't write.
Retry strategies, rollback protocols, circuit breakers for delegations, escalation chains, and failure budgets. Use when verification fails, a fix doesn't work, a sub-agent returns weak output, or you are stuck in a failure loop.
Production incident triage, rollback procedures, blast radius assessment, severity classification, and post-mortem templates. Use when production is down, a deploy broke something, users are impacted, or you need to stabilize before fixing root cause.