testing · rsmdt — Claude Code skill: write tests, debug failures, manage flaky
testing — a skill for writing effective tests: layer-specific mocking rules, failure debugging, flaky test management.
MIT ★ 587 updated 2026-10-26
/plugin marketplace add rsmdt/the-startup
/plugin install team@the-startupWhat testing does
The skill teaches the agent to write tests at the right pyramid layer. Units (60–70%) verify business logic, validation, edge cases — only external boundaries mocked: database, APIs, filesystem, time. Integration tests (20–30%) run with real database and cache, mocks only for third-party services. E2E (5–10%) have no mocking at all — critical paths like payments and auth flows.
Inside are debugging rules by failure type, a flaky test protocol (quarantine, fix in a week or delete), and a list of anti-patterns — over-mocking, testing implementation details, shared state between tests. Structure is Arrange-Act-Assert, each test verifies one observable behavior. Edge cases are mandatory: null, empty values, negative numbers, timeouts, network errors.
What it helps with
- write unit tests for validation and transformation logic
- debug integration test failure with database state
- eliminate flaky E2E tests on critical paths
- verify edge case and error coverage
- choose the right pyramid layer for new functionality
Example prompts for the agent
- Write unit tests for the discount calculation function, cover edge cases
- Debug the
rejects order when inventory insufficienttest failure - Check that payment tests run without mocks, add missing timeout checks
How to install testing
1Plugin marketplace
/plugin marketplace add rsmdt/the-startup/plugin install team@the-startupThis installs the whole team plugin with all its skills
2Skills CLI (Vercel)
npx skills add rsmdt/the-startup --skill testing -a claude-code3Manually: put the files in place
~/.claude/skills/testing/SKILL.mdCopy the whole skill folder; for one project use .claude/skills/testing/
1Skills CLI (Vercel)
npx skills add rsmdt/the-startup --skill testing -a codex2Manually: put the files in place
.agents/skills/testing/SKILL.mdFor all projects use ~/.agents/skills/
1Skills CLI (Vercel)
npx skills add rsmdt/the-startup --skill testing -a cursor2Manually: put the files in place
.cursor/skills/testing/SKILL.mdCursor also reads .agents/skills/ and .claude/skills/
1GitHub CLI
gh skill install rsmdt/the-startup plugins/team/skills/development/testing2Manually: put the files in place
.github/skills/testing/SKILL.md1Skills CLI (Vercel)
npx skills add rsmdt/the-startup --skill testing -a gemini-cli2Platform CLI
gemini skills install https://github.com/rsmdt/the-startup --path plugins/team/skills/development/testingManually: put the files in place
.devin/skills/testing/SKILL.mdThe legacy .windsurf/skills/ path still works
Claude.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 testing
How to install testing in Claude Code?
Run /plugin marketplace add rsmdt/the-startup, then /plugin install team@the-startup. Or via npx skills add rsmdt/the-startup --skill testing -a claude-code, the skill will be in ~/.claude/skills/testing/SKILL.md.
What does the testing skill do?
Teaches the agent to write tests at the right pyramid layer, mock only boundaries, debug failures, and manage flaky tests. Units have no I/O, integration tests use real databases, E2E has no mocking at all.
How does the skill determine what to mock?
Units mock database, APIs, files, time; internal classes are real. Integration tests use real database and cache, mocks only for third-party services. E2E has no mocking, everything in sandbox mode.
Source text
SKILL.md · MIT
---
name: testing
description: Writing effective tests and running them successfully. Covers layer-specific mocking rules, test design principles, debugging failures, and flaky test management. Use when writing tests, reviewing test quality, or debugging test failures.
---
## Persona
Act as a testing specialist who writes effective tests, applies layer-appropriate mocking strategies, and debugs failures systematically. You enforce test quality standards and ensure the right behavior is tested at the right layer.
**Test Context**: $ARGUMENTS
## Interface
TestDecision {
layer: Unit | Integration | E2E
mockingStrategy: string
target: string
pattern: ArrangeActAssert | GivenWhenThen
}
DebugResult {
failure: string
rootCause: string
fix: string
}
State {
context = $ARGUMENTS
scope = null
layer = null
tests = []
failures = []
}
## Constraints
**Always:**
- Test behavior, not implementation — assert on observable outcomes.
- One behavior per test — multiple assertions OK if verifying same logical outcome.
- Use descriptive test names that state the expected behavior.
- Follow Arrange-Act-Assert structure in every test.
- Mock at boundaries only — databases, APIs, file system, time.
- Use real internal collaborators — never mock application code.
- Keep tests independent — no shared mutable state between tests.
- Handle flaky tests aggressively — quarantine, fix within one week, or delete.
- Focus on business-critical paths (payments, auth, core domain logic).
- Prefer quality over quantity — 80% meaningful coverage beats 100% trivial coverage.
**Never:**
- Mock internal methods or classes — that tests the mock, not the code.
- Test implementation details — tests should survive refactoring.
- Skip edge case testing — boundaries, null, empty, negative values.
- Leave flaky tests in the main suite — they erode trust.
## Reference Materials
- [examples/test-pyramid.md](examples/test-pyramid.md) — layer-specific code examples and mocking patterns
## Workflow
### 1. Assess Scope
Identify what needs testing:
match (context) {
new feature code => write tests for new behavior
bug fix => write regression test first, then fix
refactoring => verify existing tests pass, add coverage gaps
test review => evaluate test quality and coverage
}
Determine layer distribution target:
- Unit (60-70%) — isolated business logic
- Integration (20-30%) — components with real dependencies
- E2E (5-10%) — critical user journeys
### 2. Select Layer
match (scope) {
business logic | validation | transformation | edge cases
=> Unit: mock at boundaries only, <100ms, no I/O, deterministic
database queries | API contracts | service communication | caching
=> Integration: real deps, mock external services only, <5s, clean state between tests
signup | checkout | auth flows | smoke tests
=> E2E: no mocking, real services in sandbox mode, <30s, critical paths only
}
Mocking rules by layer:
- Unit — mock external boundaries (DB, APIs, filesystem, time)
- Integration — real databases, real caches, mock only third-party services
- E2E — no mocking at all
### 3. Write Tests
Apply Arrange-Act-Assert pattern. Name tests descriptively: "rejects order when inventory insufficient"
Always test edge cases:
- Boundaries — min-1, min, min+1, max-1, max, max+1, zero, one, many
- Special values — null, empty, negative, MAX_INT, NaN, unicode, leap years, timezones
- Errors — network failures, timeouts, invalid input, unauthorized
Read examples/test-pyramid.md for layer-specific code examples.
### 4. Run Tests
Execute in order (fastest feedback first):
1. Lint/typecheck
2. Unit tests
3. Integration tests
4. E2E tests
### 5. Debug Failures
match (layer) {
Unit => {
1. Read the assertion message carefully
2. Check test setup (Arrange section)
3. Run in isolation to rule out state leakage
4. Add logging to trace execution path
}
Integration => {
1. Check database state before/after
2. Verify mocks configured correctly
3. Look for race conditions or timing issues
4. Check transaction/rollback behavior
}
E2E => {
1. Check screenshots/videos
2. Verify selectors still match the UI
3. Add explicit waits for async operations
4. Run locally with visible browser
5. Compare CI environment to local
}
}
Flaky test protocol:
1. Quarantine — move to separate suite immediately
2. Fix within 1 week — or delete
3. Common causes: shared state, time-dependent logic, race conditions, non-deterministic ordering
Anti-patterns to flag:
- Over-mocking — testing mocks instead of code
- Implementation test — breaks on refactoring
- Shared state — test order affects results
- Test duplication — use parameterized tests instead
© Rudolf S. · MIT · rsmdt/the-startup
Similar skills
All skills for Testing and QA →testing by other authors comes first.