agent-tdd-guide

Write a failing test describing the expected behavior first, run it to confirm it fails, then write the minimal implementation to make it pass. Finally refactor while keeping tests green and verify coverage reaches 80%.

Red means writing a failing test, Green means writing the minimal code to pass it, and Refactor means improving names and removing duplication while tests stay green. This skill walks you through each phase with verification commands.

This workflow requires 80% or higher coverage across branches, functions, lines, and statements, verified with a coverage command such as npm run test:coverage. Release-critical paths additionally target pass^3 eval stability before merge.

Tests must cover null and undefined inputs, empty arrays and strings, invalid types, boundary values, error paths like network failures, race conditions, large datasets, and special characters including Unicode and SQL characters.

Avoid testing internal implementation details instead of behavior, tests that depend on shared state, weak assertions that verify nothing, and failing to mock external dependencies like databases, Redis, or third-party APIs.