Two plain SKILL.md files give any AI coding agent a structured workflow to design and run claim-driven distributed-systems tests, producing a Markdown plan and findings report with 9-state verdicts.
Key Takeaways
The design skill extracts product claims, generates falsification hypotheses, and emits a §0-§9 plan including coverage adequacy argument and confidence statement.
For consistency-critical scenarios, each §7.M block binds an abstract model (register, queue, log, lock, lease, ledger) to an operation-history schema, a named checker (Porcupine linearizability, Elle serializability, exactly-once), and a nemesis with observable landing evidence.
The execute skill discovers existing tests and runbooks before inventing new harnesses, then assigns blame to SUT, harness, checker, or environment on every FAIL.
Works with Claude Code, Codex, Copilot CLI, Cursor, or Gemini; install is one paste of a URL pointing to an idempotent INSTALL.md.
