SkillVerdict
All verdicts

tdd

Install mattpocock/skills

Red-green loop rules, agreed test seams and named anti-patterns like tautological tests. Instructions only, MIT, and it asks before testing anything.

Verified · Repo on GitHub

What it does

A reference for the red → green loop that aims at tests worth keeping: test behaviour through public interfaces, agree the "seams" with you before writing any test, work in vertical slices (one test, one implementation), and avoid named anti-patterns — implementation-coupled tests, tautological assertions that recompute the expected value, and writing all tests first. Refactoring is left to review, not the loop.

Who made it / license / last checked

Matt Pocock (mattpocock/skills) · License: MIT (repo LICENSE) · Repo last commit checked: 2026-10-29

Install

claude plugin install mattpocock-skills@claude-plugins-official

From the repo README (Claude Code route; it installs the whole pack, so the skills this one calls come with it). For other agents the README gives npx skills@latest add mattpocock/skills -a <agent>. We did not run the installer.

What files / scripts it runs

SKILL.md (3,629 bytes), tests.md and mocking.md (examples and mocking rules) and an agents/openai.yaml display file. No scripts. Unlike grill-me, the model can invoke it on its own when you mention TDD or red-green-refactor.

Pros

  • Short and specific; the tautological-test rule alone catches a common AI failure
  • Confirms test seams with you first instead of testing everything
  • No code execution of its own

Cons

  • It runs your test suite through the agent's normal tools — that is the point, but it is still execution
  • Points to codebase-design and a code-review skill for vocabulary and refactoring: best with the pack

Alternatives

obra/superpowers test-driven-development (not reviewed here); a plain "write the failing test first" instruction in your project rules.

Notes

Skills.sh listed tdd at 1.2M installs (#9) on 30 Oct 2026.

Sources

Verified on 2026-11-04 through the GitHub API: it has had no commit since the 30 Oct check, so the files, sizes and verdict above still hold. Previous check on 2026-10-30, from a fresh clone of the repo (commit a496f55, 2026-10-29) and the skills.sh leaderboard fetched that day. Repo last-commit is a repo date, not a per-skill release date. Install counts are distribution, not quality.