Skip to content
Logo

Agent setup

The preferred user prompt is:

Read https://bloom.directory/SKILL.md and set up Bloom.

An agent should then:

  1. Install or build the bloom binary.
  2. Run bloom init.
  3. Mount Bloom at /bloom or the configured mount path.
  4. Inspect /bloom/docs/README.md, /bloom/status/, and the top-level tree before taking actions.
  5. Explain what it can do and ask before confirming any transaction.

Rules for agents

  • Do not write or use custom signing or broadcast code.
  • Read generated plans before asking the user to approve anything.
  • Treat missing config, disabled broadcasts, and policy failures as safety signals, not obstacles to bypass.
  • Keep live mainnet and L2 sends opt-in.

Useful discovery commands

ls /bloom/
cat /bloom/docs/README.md
cat /bloom/status/daemon.json
cat /bloom/status/backends/summary.json
ls /bloom/chains
ls /bloom/tools