Why does agent-skills' TDD skill hardcode npm test instead of using my project's test runner?

Question

When I use the test-driven-development skill from addyosmani/agent-skills on a non-JavaScript/TypeScript project, the verification checklist tells the agent to run npm test and the linked testing-patterns.md reference only covers Jest, React Testing Library, Supertest, and Playwright. Is the skill actually JS/TS-only despite claiming to be universal, and is there a fix or workaround?

Originally asked on GitHub: https://github.com/addyosmani/agent-skills/issues/404

AgentArmor team · Accepted Answer

Still open as of September 2026. Maintainers confirmed the problem and agreed on a fix direction, but no PR has landed yet.

What is wrong: `test-driven-development/SKILL.md` hardcodes `npm test` in its verification checklist, and `references/testing-patterns.md` is entirely Jest / React Testing Library / Supertest / Playwright while describing itself as covering patterns “across the stack.” A maintainer confirmed this means an agent in a Java, Go, or Rust project gets commands that do not apply.

Agreed fix direction: keep one TDD skill, make it discover the repository first (language, build system, checked-in wrappers, test runner, existing conventions), then run RED-GREEN-REFACTOR against the project's own focused-test and full-suite commands instead of a hardcoded one. The reference is to be relabeled in place as JavaScript/TypeScript examples of universal principles.

One contributor also grepped the whole catalog and found nine more npm-family commands in verification checkboxes across `incremental-implementation`, `security-and-hardening`, `planning-and-task-breakdown`, and `shipping-and-launch`; 15 of 24 skills mention npm-family commands somewhere. The maintainer agreed that makes this a catalog-wide ecosystem assumption and planned a systematic sweep rather than patching one skill.

General advice until it lands: treat `testing-patterns.md` as JS/TS-only guidance and replace `npm test` with your project's real commands. For Java, the reporter pointed to the seed4j repository (https://github.com/seed4j/seed4j) as practitioner-validated test examples; the maintainer called it a useful source, though it has not been merged into the skill.

AgentArmor team · Answer

**What a GitHub user tried on GitHub** (Outcome unclear)

The issue proposes making the TDD core ecosystem-neutral: require the agent to discover language, build system, wrappers, test runner, and conventions before choosing commands, and replace hardcoded `npm test` with the repo's own commands. Not implemented yet.

Source: https://github.com/addyosmani/agent-skills/issues/404

AgentArmor team · Answer

**What @federicobartoli tried on GitHub** (Outcome unclear)

A collaborator grepped the catalog and found nine more npm-family commands in verification checkboxes across four other skills, plus 15 of 24 skills mentioning npm-family commands and 53 TypeScript code fences versus one Python. The maintainer agreed this makes the fix a systematic sweep, but no implementation has landed.

Source: https://github.com/addyosmani/agent-skills/issues/404#issuecomment-5010846273

AgentArmor team · Answer

**What @renanfranca tried on GitHub** (Outcome unclear)

Renan França suggested seed4j (https://github.com/seed4j/seed4j) as a Java repository with carefully written test examples, which they also use locally as a reference for Codex CLI. The maintainer agreed it is a useful practitioner-validated source, but it is not merged into the skill.

Source: https://github.com/addyosmani/agent-skills/issues/404#issuecomment-4993176363

Why does agent-skills' TDD skill hardcode npm test instead of using my project's test runner?

4 answers

Add your answer or comment

Why does agent-skills' TDD skill hardcode npm test instead of using my project's test runner?

4 answers

Add your answer or comment

Related questions