maida check

Check the latest Claude Code task captured in this initialized repository, then open the same task using the printed viewer command. This command is available from v0.6.1.

maida check

When Maida is installed in your project environment, use uv run maida check. The report prints the trace ID and uv run maida view TRACE_ID in that case; otherwise it prints maida view TRACE_ID.

The built-in checks require successful completion, no observed loop warnings and no recorded guardrail events. No policy or baseline is needed, and existing policy files do not affect these first-run checks. The report covers observed tool activity and session lifecycle, not answer correctness or complete model-call, token or latency coverage.

The newest session must have finished and contain useful evidence. Missing, incomplete or broken capture produces a concrete recovery action; the command never substitutes SDK runs or an older completed task for unfinished new work. Run maida init if capture is absent or detached. Saved Claude tasks remain readable by explicit trace ID after detach.

--format json and --format markdown produce machine-readable or shareable reports; the trace ID and viewer command go to stderr so redirected report output stays clean. Human-readable text is the default and includes both on stdout.

This command does not change maida assert, bare maida view or other SDK/Python read defaults. Use maida assert TRACE_ID --baseline PATH for an explicit captured task’s baseline gate.

Exit codes: 0 the three checks passed; 1 an observed check failed; 2 setup, capture or format needs repair; 10 an unexpected internal failure. A failing report still prints the selected task and its viewer command.