AI features you ship to users: structured output, tool schemas, prompt injection, evals. Use when "the model returns bad JSON", "it hallucinates", "stop it calling the wrong tool", "add evals", or an LLM feature can trigger refunds, emails or writes. Covers schema-constrained output, idempotent tool calls, confirmation gates. For agents editing your repo use agent-guardrails.
安装
npx skills add rainmanjam/poka-yoke --skill llmnpx skills add rainmanjam/poka-yoke --skill llm -a claude-code -gnpx skills add rainmanjam/poka-yoke --skill llm -a codex -ggemini skills install https://github.com/rainmanjam/poka-yoke.git --path plugins/poka-yoke/skills/llmgh skill install rainmanjam/poka-yoke llmgit clone --depth 1 https://github.com/rainmanjam/poka-yoke /tmp/rainmanjam__poka-yoke
mkdir -p ~/.claude/skills ~/.agents/skills
cp -R /tmp/rainmanjam__poka-yoke/plugins/poka-yoke/skills/llm ~/.claude/skills/llm # Claude Code
cp -R /tmp/rainmanjam__poka-yoke/plugins/poka-yoke/skills/llm ~/.agents/skills/llm # Codex, Cursor, Gemini CLI, Copilot, OpenCode