tdd: Red, Green, Refactor for Agentic Coding

Install this skill:

npx skills add mattpocock/skills skill=tdd -y -g

Source: mattpocock/skills/tdd

What it does

tdd guides an agent through test-driven development with a red, green, refactor loop.

The central rule is that tests should verify behavior through public interfaces, not implementation details.

Good tests

Good tests read like specifications. They describe what the system does from the outside.

They survive refactors because they do not care about private functions, internal collaborators, or the exact shape of implementation code.

Bad tests

Bad tests couple themselves to implementation details. They mock internals, test private methods, or fail when code is refactored even though behavior did not change.

The skill pushes the agent away from those tests.

One behavior at a time

tdd rejects horizontal TDD:

WRONG:
RED: test1, test2, test3, test4
GREEN: impl1, impl2, impl3, impl4

Instead, it works vertically:

RIGHT:
RED → GREEN: test1 → impl1
RED → GREEN: test2 → impl2
RED → GREEN: test3 → impl3
REFACTOR: clean up after green

This keeps the agent from writing tests for imagined behavior before it understands the implementation.

How it fits the workflow

grill-me → domain-model → to-prd → to-issues → tdd

Use tdd after to-issues has produced a vertical slice. Then implement one behavior at a time through public interfaces.

Pairs well with

• to-issues, for vertical-slice GitHub issues
• to-prd, for the source PRD
• domain-model, when tests need stable domain language