Skip to content
Agentmart
Submit a tool

tdd

Use when implementing or fixing production code with tests.

Skill by ouonet

IndexedMIT★ 12
Write a review

Install

Works with Claude Code, Claude apps, Codex, Cursor, Gemini CLI, GitHub Copilot, VS Code, OpenCode, Goose, Windsurf. Read the repository before you give any tool access to your data.

npx agentmarket install skill/ouonet-tdd

Asks which of your tools to add it to (Claude Code, Codex, Cursor, VS Code, Windsurf, Gemini CLI, Claude Desktop) and shows the change first. See https://agentmart.dev/cli.

Get the weekly diff

The weekly diff: one email a week with new agent tools, notable updates and grade changes, plus security alerts for tools you follow. Double opt-in, one-click unsubscribe.

We use your email only for this newsletter. Privacy

About

Use when implementing or fixing production code with tests.

Use cases: testing

No reviews yet. Installed it? Reviews come from verified installers only. How reviews work

Permissions

What this skill can do and what it asks for, from its SKILL.md. Checked Oct 25, 2026.

Tools
No tools found
Secrets
None asked for
Install scripts
None found

Network hosts it talks to 1

  • codeload.github.com · found in install command

Files it touches

  • ~/.claude/skills/tdd

What's inside

Skill name
tdd
Description
Use when implementing or fixing production code with tests.
License
MIT
Instructions
46 lines in SKILL.md · view on GitHub

Files (1)

  • SKILL.md

SKILL.md (first lines)

# TDD

If `topology=multi-module` (triage announcement or coordinator spec/plan declaration), read `../references/multi-module.md`. Run implementation and local acceptance inside the task's registered repository/module. Update that module's spec and plan; update the coordinator only for shared-contract reality or workspace state.

## Acceptance mode

Read the current plan task's `acceptance:` (or equivalent).

**Automated acceptance** (test, command, or script that must fail before the change and pass after):
- **No production code without a failing check first.** Wrote code before the failing check? Delete it. Rewrite from the check.
- RED (fail for the *right reason*) → GREEN (minimum to pass) → refactor → **sync docs** → commit → flip the plan checkbox.

**Manual acceptance** (explicit steps + expected result in the plan — config, docs, infra, prompts, content, or other non-testable deliverables):
- Do **not** invent an automated test solely to satisfy RED-GREEN.
- Perform the manual check; keep the change minimal; run any declared mechanical checks; **sync docs**; commit when appropriate; flip the plan checkbox.

**No acceptance at all** (prototypes, generated code, throwaway scripts only): ask the user before writing production code.

## After green / manual pass

**edit `docs/staging/plans/YYYY-MM-DD-<topic>.md` and change this task's `- [ ]` to `- [x]`. Do not start the next task without this edit.**

**Sync docs** means:
- If staging spec exists (`docs/staging/specs/*.md`): update it to match code reality.
- If no staging spec (small task): update living docs (README, tech-spec, comments) directly.

When local module tasks are complete and green, run the coordinator's cross-module acceptance against the current workspace contents before handing off to `ship`.

For a linked multi-repository change, defer commits until integration passes, then follow the commit protocol in `../references/multi-module.md`. This is the exception to the per-cycle commit order above. If any commit fails, stop at `partial-commit`; never rewrite or discard completed commits automatically.

## Refactor

Passing tests are not a quality bar.

`<gate>` Before committing: (1) run lint/format/typecheck — green if declared; (2) evaluate against `convention`; (3) assess design — coupling, cohesion, abstraction, idiomatic for the stack. State what you ran, assessed, and changed — or why not. See `../references/quality.md`. `</gate>`

## Don't
- Test passes without the impl (tests nothing).
- Mock the unit under test (tests the mock).
- Assert many behaviors in one test (split).
- Skip "watch it fail" when acceptance is automated (you don't know what it tests).
- Edit the test to match buggy code (tests the bug).
- Add abstractions not required by the current acceptance (GREEN phase — not refactor).
- Edit files outside the failing check's scope.
- Create ad-hoc summary, notes, or analysis files not defined in the plan or required by a loaded skill.

From ouonet/praxis, MIT.

More from ouonet/praxis

Signals

Stars
12
Forks
1
Open issues
0
Last commit (this folder)
Aug 4, 2026 (2 months ago)

From GitHub (fetched Oct 24, 2026).

Details

License
MIT
First seen
Oct 24, 2026
Last checked
Oct 24, 2026

More from ouonet

Cite this page

tdd skill. Agent Mart. https://agentmart.dev/skills/ouonet-tdd (last updated 2026-10-24)