Audit or rewrite LLM instructions — a skill (SKILL.md), CLAUDE.md, agent definition, system prompt, tool description or a single prompt — against Anthropic's prompting guidance for current Claude models. Diagnoses first, then proposes the rewrite with the reason for each change; it does not overwrite the file until the user approves.
Install
npx skills add orrrrli/clawd-harness --skill prompt-rewritenpx skills add orrrrli/clawd-harness --skill prompt-rewrite -a claude-code -gnpx skills add orrrrli/clawd-harness --skill prompt-rewrite -a codex -ggemini skills install https://github.com/orrrrli/clawd-harness.git --path skills/engineering/prompt-rewritegh skill install orrrrli/clawd-harness prompt-rewritegit clone --depth 1 https://github.com/orrrrli/clawd-harness /tmp/orrrrli__clawd-harness
mkdir -p ~/.claude/skills ~/.agents/skills
cp -R /tmp/orrrrli__clawd-harness/skills/engineering/prompt-rewrite ~/.claude/skills/prompt-rewrite # Claude Code
cp -R /tmp/orrrrli__clawd-harness/skills/engineering/prompt-rewrite ~/.agents/skills/prompt-rewrite # Codex, Cursor, Gemini CLI, Copilot, OpenCode