Adversarial validation for all stages. Detects fabrications, identifies analytical flaws, challenges assumptions, makes approval decisions. Seeks problems rather than confirming quality. Does NOT complete deliverables or fix issues.
安装
npx skills add ensingm2/AI-threat-modeling-rulesets --skill quality-criticnpx skills add ensingm2/AI-threat-modeling-rulesets --skill quality-critic -a claude-code -gnpx skills add ensingm2/AI-threat-modeling-rulesets --skill quality-critic -a codex -ggemini skills install https://github.com/ensingm2/AI-threat-modeling-rulesets.git --path .ai-instructions/skills/quality-criticgh skill install ensingm2/AI-threat-modeling-rulesets quality-criticgit clone --depth 1 https://github.com/ensingm2/AI-threat-modeling-rulesets /tmp/ensingm2__AI-threat-modeling-rulesets
mkdir -p ~/.claude/skills ~/.agents/skills
cp -R /tmp/ensingm2__AI-threat-modeling-rulesets/.ai-instructions/skills/quality-critic ~/.claude/skills/quality-critic # Claude Code
cp -R /tmp/ensingm2__AI-threat-modeling-rulesets/.ai-instructions/skills/quality-critic ~/.agents/skills/quality-critic # Codex, Cursor, Gemini CLI, Copilot, OpenCode