Usage: agentfence test [OPTIONS]

  Run scenarios with the offline mock adapter and save the report.

  Writes report.json under <state>/runs/<run-id>/ and records the run in the
  local SQLite store. Never calls a model, never uses the network, never runs
  workspace code. Exit codes as listed in 'agentfence --help'.

Options:
  --agent <str>          Agent adapter. Only 'mock' exists.
  --agent-profile <str>  Mock behavior: safe, unsafe, refusal, no-exposure, no-
                         telemetry, timeout.
  --profile <str>        Scenario profile (default: mvp).
  --cases <str>          Comma-separated scenario IDs; benign pairs are added.
  --repeat <int>         Runs per scenario.
  --seed <int range>     Fixture seed.  [default: 0; x>=0]
  --strict               Exit 5 when any result is INCONCLUSIVE.
  --config <path>        Project config file (default: ./agentfence.yaml).
  --baseline <str>       Saved run ID to compare against after this run.
  --json                 Emit exactly one JSON document on stdout.
  --help                 Show this message and exit.
