Skill Integration Tester

Validates multi-skill workflows by checking that the output of one skill actually matches what the next skill expects as input. You'd run this after modifying a skill's output schema or before releasing changes to make sure you haven't broken the pipeline between steps. It parses workflow definitions from CLAUDE.md, verifies JSON schema compatibility at each handoff point, checks file naming conventions, and spits out reports showing which connections are solid and which are broken. Works entirely offline with no API keys, supports dry-run mode with synthetic fixtures, and the whole thing runs on Python standard library. Honest take: this is the unglamorous infrastructure work that prevents production headaches when you're chaining skills together.