new-feature — Claude Code skill: TDD feature workflow
Runs a new feature through four ordered steps: failing tests first, then code, then the full suite. Handy when regressions are the thing you keep missing.
npx skills add undeadlist/claude-code-agents --skill new-feature -a claude-codeWhat new-feature does
The skill splits feature work into a fixed sequence. It hands your requirements to the test-writer subagent, which reads the spec and writes failing tests covering the happy path and edge cases. Then code-fixer picks up the requirements plus those test files, builds the feature to match the code already in the repo, and gets the new tests green.
Verification comes next: test-runner executes everything, new tests and old, so regressions surface. UI features get a fourth pass, where browser-qa-agent opens the feature, exercises the interactions and watches the console for errors. Nothing runs in parallel here.
The author advises skipping the browser step for pure API work, library or utility code, database migrations and background jobs. Requirements are mandatory at invocation — with no spec there is nothing to test against. Available for Claude, Codex, Cursor, GitHub Copilot, Gemini CLI and OpenCode.
What it helps with
- start a feature from tests instead of code
- cover edge cases before implementation
- catch regressions in the existing suite
- exercise a new UI in the browser
- break feature work into ordered steps
Example prompts for the agent
/new-featureFeature: user profile update endpoint — PUT /api/users/:id, allow name, email, avatar, require auth, validate email format/new-featureadd a date filter to the orders list and cover the edge cases/new-featurea CSV parsing utility function, skip browser QA
How to install new-feature
1Skills CLI (Vercel)
npx skills add undeadlist/claude-code-agents --skill new-feature -a claude-code2Manually: put the files in place
~/.claude/skills/new-feature/SKILL.mdCopy the whole skill folder; for one project use .claude/skills/new-feature/
1Skills CLI (Vercel)
npx skills add undeadlist/claude-code-agents --skill new-feature -a codex2Manually: put the files in place
.agents/skills/new-feature/SKILL.mdFor all projects use ~/.agents/skills/
1Skills CLI (Vercel)
npx skills add undeadlist/claude-code-agents --skill new-feature -a cursor2Manually: put the files in place
.cursor/skills/new-feature/SKILL.mdCursor also reads .agents/skills/ and .claude/skills/
1GitHub CLI
gh skill install undeadlist/claude-code-agents new-feature2Manually: put the files in place
.github/skills/new-feature/SKILL.md1Skills CLI (Vercel)
npx skills add undeadlist/claude-code-agents --skill new-feature -a gemini-cli2Platform CLI
gemini skills install https://github.com/undeadlist/claude-code-agents --path skills/new-featureManually: put the files in place
.devin/skills/new-feature/SKILL.mdThe legacy .windsurf/skills/ path still works
Skills CLI (Vercel)
npx skills add undeadlist/claude-code-agents --skill new-feature -a opencodeClaude.ai
Zip the folder and upload it in Settings → Capabilities → Skills
Step by step, with commands for each platform: How to install skills in Claude Code: five ways
Questions about new-feature
How do I install new-feature in Claude Code?
Run npx skills add undeadlist/claude-code-agents --skill new-feature -a claude-code. The skill file lands at ~/.claude/skills/new-feature/SKILL.md.
What does new-feature do?
It drives a feature through TDD: test-writer produces failing tests, code-fixer implements the code, test-runner runs the whole suite. UI features also get a browser pass via browser-qa-agent.
When can I skip the browser QA step?
For features with no interface: pure API work, library or utility code, database migrations and background jobs. Those cases stop after the test run.
Source text
SKILL.md · MIT
--- name: new-feature description: Develop new features with test-first approach (TDD) --- # New Feature Workflow Build features using test-driven development. Provide the feature specification when invoking. ## Execution (Sequential) ### Step 1: Write Tests Spawn `test-writer` with the feature requirements: - Analyze feature specification - Write failing tests first - Cover happy path + edge cases ### Step 2: Implement Feature Spawn `code-fixer` with the feature requirements and test files: - Implement feature following existing patterns - Make all new tests pass ### Step 3: Verify Spawn `test-runner`: - Run all tests (new + existing) - Verify no regressions ### Step 4: Browser QA (if UI feature) Spawn `browser-qa-agent`: - Navigate to the new feature - Test interactions - Check for console errors ## Usage Provide feature requirements when invoking: ``` /new-feature Feature: User profile update endpoint - PUT /api/users/:id - Allow updating: name, email, avatar - Require authentication - Validate email format - Return updated user ``` ## Skip browser-qa for - Pure API features (no UI) - Library/utility code - Database migrations - Background jobs
© undeadlist · MIT · undeadlist/claude-code-agents
