What problem does it solve? Projection errors in FutureFin are silent — numbers look plausible but are wrong — so eyeballing a chart is never acceptance. This Skill tells you exactly what counts as proof for a change: which of the three test suites to run, how the TestApp integration harness works, and how to write regression, parity, and context-field tests that actually catch bugs. ## Core Features & Use Cases - Evidence standards table: Maps each change type (refactor, model change, bug fix, visual change, absence assertion, context field) to the required proof, such as bit-exact regression capture or predict-then-measure. - Full test inventory and invocation: Documents the engine unit tests (cargo), backend integration tests against Postgres via TEST_DATABASE_URL, and frontend Vitest suites, including the isolated_pool() per-test schema mechanics and cleanup commands. - Harness knobs and house style: Explains TestConfig/spawn_with options (trusted proxy auth, base path, SPA fallback, MCP kill-switch) and the fake-behind-a-trait pattern for outbound integrations like the HA identity provider. - Use Case: Before merging a change to the FIRE projection math, use this Skill to write a failing pinning test first, run the full workspace suite against the ff-test-db container on port 5433, and verify the fire-parity.json fixture still matches on both client and server. ## Quick Start Ask the AI to determine which tests must run locally and what evidence is required before merging your FutureFin change.