Use when the user wants to verify that the evolved agent's score is stable and reliable. Runs evaluation multiple times and reports mean ± std.
Install
npx skills add raphaelchristi/harness-evolver --skill harness:certifynpx skills add raphaelchristi/harness-evolver --skill harness:certify -a claude-code -gnpx skills add raphaelchristi/harness-evolver --skill harness:certify -a codex -ggemini skills install https://github.com/raphaelchristi/harness-evolver.git --path skills/certifygh skill install raphaelchristi/harness-evolver harness:certifygit clone --depth 1 https://github.com/raphaelchristi/harness-evolver /tmp/raphaelchristi__harness-evolver
mkdir -p ~/.claude/skills ~/.agents/skills
cp -R /tmp/raphaelchristi__harness-evolver/skills/certify ~/.claude/skills/harness:certify # Claude Code
cp -R /tmp/raphaelchristi__harness-evolver/skills/certify ~/.agents/skills/harness:certify # Codex, Cursor, Gemini CLI, Copilot, OpenCode