Usage: agentfence [OPTIONS] COMMAND [ARGS]...

  AgentFence Lab: evidence-first security regression testing for AI coding
  agents.

  Pre-alpha. Nothing here calls a model, launches a real agent or uses the
  network. 'test' runs the offline mock adapter only.

  Exit codes:
    0  success (may include INCONCLUSIVE unless --strict)
    1  one or more observed boundary FAILs
    2  usage, configuration or validation error
    3  isolation or security preflight failure
    4  infrastructure failure or timeout
    5  INCONCLUSIVE result under --strict
    6  unsupported target, agent or configuration
    130  cancelled with Ctrl-C (nothing saved, workspaces removed)

Options:
  --help  Show this message and exit.

Commands:
  version    Print the AgentFence Lab version.
  doctor     Check host prerequisites and validate configuration.
  test       Run scenarios with the offline mock adapter and save the report.
  cleanup    Delete saved runs older than the retention period.
  ui         Serve the read-only dashboard and API on loopback.
  report     Export a saved run's report after verifying it.
  scenarios  Inspect bundled scenarios.
  runs       Inspect saved runs.
