testwrite

A Claude Code skill that generates a real test suite for new or changed code. Covers happy path, boundaries, failure modes, abuse cases, idempotency, and regression anchors. Matches your project's existing test framework + conventions.

Categories covered

1. Happy path — does it do the thing
2. Input boundaries — min, max, edge values, null/undefined, wrong type
3. Failure modes — dependency down, malformed input, empty results
4. Security & abuse — SQL injection, prompt injection, path traversal, oversized input, unicode
5. Idempotency & side effects — runs once on success, doesn't fire on failure
6. Regression anchors — bug-fix tests, "old behavior still works" tests

Install

git clone https://github.com/matthewkim323/testwrite ~/.claude/skills/testwrite

Use

testwrite src/auth.ts              # all public functions in that file
testwrite src/auth.ts:validateJWT  # one specific function
testwrite --diff                   # tests for everything in git diff HEAD
testwrite --minimal                # fast: happy + 1 failure + 1 boundary per fn
testwrite --comprehensive          # all categories + property-based fuzzing
testwrite --integration            # integration tests instead of unit (skip mocks)
testwrite --framework pytest       # override auto-detection

Or just say "write tests for X" / "test this file" / "cover this with tests".

Pairs with other skills

• Run before /prodcheck — CHECK 7 in prodcheck flags missing test coverage. testwrite fixes that.
• Run before /prwrite — your PR description's Testing section will read honestly because you actually wrote tests.

License

MIT.