What problem does it solve? Investigating failures, slow functions, and deploy regressions in a running Convex app requires knowing the right log filters and tool limits, and raw log dumps are hard to act on. This Skill turns natural-language operational questions into scoped, evidence-backed queries against the official Convex MCP read tools. ## Core Features & Use Cases - Failures View: Fetches a bounded log window and groups errors by function and message, returning counts plus a representative stack trace. - Health View: Surfaces typed 72-hour read-limit and OCC events from the insights tool, routing perf and cost root-causing to the advisor skill. - Trace View: Filters logs client-side to a single requestId or function to explain why one specific call failed. - Use Case: After a deploy, ask whether the new version broke anything; the Skill compares failure onset timestamps against the deployment version and returns a dashboard deep link for one-click verification. ## Quick Start Ask the agent to check whether any Convex functions started failing after the latest deploy and show the evidence with a dashboard link.
