Use when the user asks for consequential knowledge, assumption, implementation, or verification gaps in accessible chats and current projects, including architecture, technology selection, completed changes, post-launch operations, and peer practices. Also use a lightweight pass at consequential decision checkpoints when a specific unexamined assumption could change the recommendation.
安装
npx skills add host452b/polish --skill find-blind-spotsnpx skills add host452b/polish --skill find-blind-spots -a claude-code -gnpx skills add host452b/polish --skill find-blind-spots -a codex -ggemini skills install https://github.com/host452b/polish.git --path skills/find-blind-spotsgh skill install host452b/polish find-blind-spotsgit clone --depth 1 https://github.com/host452b/polish /tmp/host452b__polish
mkdir -p ~/.claude/skills ~/.agents/skills
cp -R /tmp/host452b__polish/skills/find-blind-spots ~/.claude/skills/find-blind-spots # Claude Code
cp -R /tmp/host452b__polish/skills/find-blind-spots ~/.agents/skills/find-blind-spots # Codex, Cursor, Gemini CLI, Copilot, OpenCode