Blitz Reviewer

Audit a Blitz integration and report what's healthy, what's stale, and what's wrong — across the environment (MCP, SDK, skills, API key) and the user's own code (methods, bodies, enums).

The Blitz MCP is the source of truth. Never judge a call against a memorized or bundled API spec — ask the MCP for the live endpoint, request body, and enum values and compare. That's why the MCP is checked first and pushed hard: it makes every other check accurate. Output is a pass/warn/fail checklist; nothing changes until the user confirms each fix.

This is diagnosis, not a refactor. Run the read-only checks, collect findings into one report, then work a remediation queue one confirmed fix at a time. Spend no credits doing it — key-info is free; never fire a search or enrichment to "test."

Quick start

Work the checks in order (MCP first — it's what makes checks 4–6 authoritative). Collect every finding, then present one checklist and a confirmation-gated remediation queue.