Hatchet CLI Agent Skills

This skill package teaches AI agents how to use the Hatchet CLI to manage workflows, workers, and runs.

When to use these skills

Read the relevant reference document before performing any Hatchet CLI task:

• Setting up the CLI or creating a profile → references/setup-cli.md
• Running Hatchet locally without a token (embedded mode) → references/local-dev-embedded.md
• Starting a worker → references/start-worker.md
• Triggering a workflow and waiting for results → references/trigger-and-watch.md
• Debugging a failed or stuck run → references/debug-run.md
• Replaying a run with the same or new input → references/replay-run.md

Key conventions

• For local development, prefer embedded mode: no token, account, or server needed (references/local-dev-embedded.md).
• Always specify a profile with -p HATCHET_PROFILE unless a default profile is set. Embedded mode needs no profile.
• Use -o json for machine-readable output when parsing responses.
• Write workflow input to a temp file (e.g. /tmp/hatchet-input-$(date +%s)-$$.json) to avoid collisions.
• Clean up temp files after use.