Bug Fix TDD

Reproduce bugs with a failing test, then apply the minimum fix. This skill is used by the automated bug-fix agent in CI but can also be invoked manually.

TDD Workflow

Phase 1 — Analysis & Failing Test (Red)

1. Parse the bug report: extract description, steps to reproduce, expected vs actual behavior
2. Find relevant code: use Grep/Glob to locate the component, hook, or route mentioned in the bug
3. Write a unit test that reproduces the bug — the test MUST FAIL
4. Run the test: pnpm run test:nonInteractive -- <test-file-path>
5. Verify failure reason: the test must fail because of the bug, not because of import errors or unrelated issues
6. Retry if needed: if the test passes (bug not reproduced), try a different approach (max 3 attempts)
7. Write bug-analysis.md with findings (see format below)

Constraints: Do NOT modify source files in Phase 1. Only create/edit test files and bug-analysis.md.

Phase 2 — Fix (Green)

1. Read the failing test and bug-analysis.md
2. Apply the MINIMUM fix to make the test pass — do not over-engineer
3. Run the single test: pnpm run test:nonInteractive -- <test-file-path>
4. Run the full suite: pnpm run test:nonInteractive
5. Run static checks: pnpm run lint and pnpm run type-check
6. Retry if needed: if any check fails, adjust the fix (max 5 attempts)
7. Write pr-body.md and fix-title.txt

Constraints: Do NOT run git, gh, or modify .env files.

Phase 2b — Direct Fix (Fallback)

If Phase 1 cannot reproduce the bug in a test (test passes after 3 attempts), Phase 2b runs instead of Phase 2.

1. Read bug-analysis.md and issue-body.md for context
2. Apply the MINIMUM fix based on code analysis alone
3. If you CAN write a regression test, do so — but it is not required
4. Run the full suite: pnpm run test:nonInteractive
5. Run static checks: pnpm run lint and pnpm run type-check
6. Retry if needed: if any check fails, adjust the fix (max 5 attempts)
7. Write pr-body.md and fix-title.txt — note in the PR body that no regression test was possible

Constraints: Same as Phase 2. Do NOT run git, gh, or modify .env files.

Test Placement Rules

• Tests go in __tests__/ directories colocated with the source file
• If a test file already exists for the component, add a new describe('Bug #N', ...) block instead of creating a new file
• Naming: <component-name>.test.tsx or <hook-name>.test.ts
• Example: source at renderer/src/features/skills/components/card-skill.tsx → test at renderer/src/features/skills/components/__tests__/card-skill.test.tsx

Test Patterns

Component test (simplest)

import { render, screen, waitFor } from '@testing-library/react'
import userEvent from '@testing-library/user-event'
import { QueryClient, QueryClientProvider } from '@tanstack/react-query'

const queryClient = new QueryClient({
  defaultOptions: { queries: { retry: false } },
})

render(
  <QueryClientProvider client={queryClient}>
    <MyComponent prop="value" />
  </QueryClientProvider>
)

await userEvent.click(screen.getByRole('button', { name: /save/i }))
await waitFor(() => {
  expect(screen.getByText('Saved')).toBeVisible()
})

Route-level component test

import { createTestRouter } from '@/common/test/create-test-router'
import { renderRoute } from '@/common/test/render-route'

const router = createTestRouter(MyPage, '/my-page')
renderRoute(router, { permissions: { canManageClients: true } })

await waitFor(() => {
  expect(screen.getByRole('heading', { name: /my page/i })).toBeVisible()
})

Hook test

import { renderHook, waitFor } from '@testing-library/react'
import { QueryClient, QueryClientProvider } from '@tanstack/react-query'

const queryClient = new QueryClient({
  defaultOptions: { queries: { retry: false } },
})
const Wrapper = ({ children }) =>
  <QueryClientProvider client={queryClient}>{children}</QueryClientProvider>

const { result } = renderHook(() => useMyHook(), { wrapper: Wrapper })

await waitFor(() => expect(result.current.isLoading).toBe(false))
expect(result.current.data).toEqual({ ... })

API mock override (return different data)

import { mockedGetApiV1BetaWorkloads } from '@mocks/fixtures/workloads/get'

mockedGetApiV1BetaWorkloads.override((data) => ({
  ...data,
  workloads: [], // Force empty state
}))

API mock error response

import { HttpResponse } from 'msw'

mockedGetApiV1BetaWorkloads.overrideHandler(() =>
  HttpResponse.json({ error: 'Server error' }, { status: 500 })
)

Request recording (for mutations)

import { recordRequests } from '@/common/mocks/node'

const rec = recordRequests()
// ... trigger action ...
const request = rec.recordedRequests.find(
  (r) => r.method === 'POST' && r.pathname === '/api/v1beta/workloads'
)
expect(request?.payload).toMatchObject({ name: 'my-server' })

bug-analysis.md Format

## Bug Summary

<1-2 sentences describing the bug>

## Root Cause

<Technical explanation of why the bug occurs>

## Relevant Files

- `path/to/source.tsx` — <what it does>
- `path/to/related.ts` — <why it's relevant>

Test file: path/to/\_\_tests\_\_/component.test.tsx

## Proposed Fix

<Description of the minimum change needed>

## Files to Modify

- `path/to/file.tsx` — <what to change>

Important: The Test file: line must be on its own line starting with exactly Test file: followed by the path. This is parsed by the CI workflow.

pr-body.md Format

## Summary

Fixes #<issue-number>.

- <1-2 bullet points describing the fix>

## Test

- Added regression test in `<test-file-path>`
- Test reproduces the bug (fails before fix, passes after)

---

_Automated fix by Claude Code TDD Agent_

fix-title.txt Format

Single line, conventional commit format:

fix(<scope>): <description> (#<issue-number>)

Example: fix(skills): prevent crash when metadata is undefined (#423)

Related Skills

• testing-with-api-mocks — Auto-generated MSW fixtures and mock basics
• testing-api-assertions — Verifying mutations with recordRequests()
• testing-api-overrides — Conditional mock responses for testing filters/params