Apify MCP Agent Evaluation Guide
Testing & QASkill Packaged as a skill by Orbit
Builds a set of test cases to check whether an AI agent's tools actually work as expected, then uses the failures to pinpoint and fix broken tool instructions.
Orbit turned public content into a SKILL.md you can read before adding it.
Checked by Orbit: nothing risky found in what it runs or reads.
What you’ll need
LangfuseStores the eval datasets, traces and scores
Or: Braintrust, LangSmith
ApifyThe MCP server being tested runs real Apify tools and Actors against your account
Or: Firecrawl, Bright Data