Root-causes AI agent failures from MLflow traces without requiring a code change, regression suite, scorer, or evaluation run. Use when the user asks to debug, investigate, explain, or determine why an agent produced an output. Prefer an existing matching trace; reproduce only when no usable trace exists.
Install
npx skills add mlflow/skills --skill debug-agentnpx skills add mlflow/skills --skill debug-agent -a claude-code -gnpx skills add mlflow/skills --skill debug-agent -a codex -ggemini skills install https://github.com/mlflow/skills.git --path debug-agentgh skill install mlflow/skills debug-agentgit clone --depth 1 https://github.com/mlflow/skills /tmp/mlflow__skills
mkdir -p ~/.claude/skills ~/.agents/skills
cp -R /tmp/mlflow__skills/debug-agent ~/.claude/skills/debug-agent # Claude Code
cp -R /tmp/mlflow__skills/debug-agent ~/.agents/skills/debug-agent # Codex, Cursor, Gemini CLI, Copilot, OpenCode