maida drift

Evaluates a completed native Maida trace window against one agent baseline without executing the agent or changing the source traces.

maida drift --window RUNS_DIR --baseline BASELINE [options]

Argument/Option	Default	Description
--window	required	Native Maida runs/ directory containing completed traces
--baseline, -b	required	Baseline JSON for one agent; directories are not yet accepted
--policy	.maida/policy.yaml	Tier-aware gate policy
--agent	baseline source_run_name	Explicit agent selector for legacy baselines
--format, -f	text	text, json, or verdict-first markdown
--json-out	-	Atomically write report schema 2.0.1 to a sidecar

maida drift --window /srv/agents/orders/runs \
  --baseline .maida/baselines/orders-agent.json \
  --policy .maida/policy.yaml \
  --format markdown --json-out orders-agent-drift.json

The baseline selects matching run_name values from a mixed-agent window. Run the command once per baseline. Exit 0 means PASS or neutral INCONCLUSIVE, 1 means FAIL, 2 means invalid input, and 10 means an internal error.

See Scheduled behavioral regression checks for sample validation, scheduler guidance, canary promotion, and planned input adapters.