Test Improvement Skill | Agent Skills

Test Improvement

Improve test quality: review existing, refactor structure, fill coverage gaps.

Use TodoWrite to track these 6 phases:

Choose mode
Explore test structure
Run coverage analysis
Review with language agent
Apply improvements
Verify & report

Phase 1: Choose Mode

$ARGUMENTS:

review → Analyze current tests, identify issues
refactor → Combine to tabular, remove duplicates, align style
coverage → Generate tests for uncovered code
full → All of the above
(empty) → Ask what to do

If no argument provided, use AskUserQuestion:

| Header | Question | Options | | ------ | ----------------------- | --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | Mode | What should I focus on? | 1. Review existing - Analyze current tests, identify issues 2. Refactor tests - Combine to tabular, remove duplicates, align style 3. Fill coverage gaps - Generate tests for uncovered code 4. Full improvement - All of the above |

Phase 2: Background Exploration (Parallel)

Spawn BOTH agents in a single message:

Task(
  subagent_type="Explore",
  run_in_background=true,
  description="Test structure scan",
  prompt="Explore test structure:
  1. Find test files: Glob for *_test.go, test_*.py, *.test.ts, *.spec.ts
  2. Identify frameworks (testify, pytest, vitest, jest)
  3. Find patterns: table-driven, parametrize, test.each usage
  4. Locate test helpers and fixtures
  5. Check mock patterns (mockery, pytest-mock, vi.mock)
  Return: language, framework, patterns found, helper locations"
)

Task(
  subagent_type="Explore",
  run_in_background=true,
  description="Coverage analysis",
  prompt="Run coverage and identify gaps:

  Go: go test -coverprofile=/tmp/claude/cov.out ./... && go tool cover -func=/tmp/claude/cov.out
  Python: pytest --cov=. --cov-report=term-missing
  TypeScript: bun test --coverage

  EXCLUDE from coverage (don't test these):
  - Test files, test helpers, fixtures
  - Generated code (*_gen.go, generated/)
  - Mock files (mocks/, mock_*.go)
  - CLI entrypoints (main.go, cmd/, __main__.py)
  - Type definitions only files

  Return: overall %, packages below 70%, uncovered business logic functions"
)

Save agent IDs for potential resumption if session is interrupted.

Phase 3: Collect Exploration Results

TaskOutput(task_id=<structure_agent_id>, block=true)
TaskOutput(task_id=<coverage_agent_id>, block=true)

Merge findings to inform next phase.

Phase 4: Spawn Language-Specific Test Agent

Based on detected language, spawn ONE appropriate agent:

Go Tests

Task(
  subagent_type="go-tests",
  description="Go test review",
  prompt="Review Go tests for quality issues.

  FOCUS ON:
  - Tests that should be table-driven (combine similar)
  - Pointless tests (trivial getters, constructors)
  - Duplicate tests (same scenario multiple ways)
  - Mock patterns (prefer mockery --with-expecter)
  - Setup duplication (extract helpers)

  Return structured findings with file:line references."
)

Python Tests

Task(
  subagent_type="py-tests",
  description="Python test review",
  prompt="Review Python tests for quality issues.

  FOCUS ON:
  - Tests that should use @pytest.mark.parametrize
  - Pointless tests (trivial behavior)
  - Duplicate tests
  - Mock patterns (pytest-mock with spec=)
  - Fixture reuse opportunities

  Return structured findings with file:line references."
)

TypeScript Tests

Task(
  subagent_type="ts-tests",
  description="TypeScript test review",
  prompt="Review TypeScript tests for quality issues.

  FOCUS ON:
  - Tests that should use test.each
  - Pointless tests (prop renders, default state)
  - Duplicate tests
  - Mock patterns (vi.fn, vi.mock, vi.mocked)
  - React Testing Library best practices

  Return structured findings with file:line references."
)

Phase 5: Apply Improvements

Based on agent findings, apply changes:

Combining Tests Pattern

| Language | Pattern | | ---------- | ------------------------------------------------ | | Go | Table-driven with t.Run(tc.name, ...) | | Python | @pytest.mark.parametrize with pytest.param() | | TypeScript | test.each([{input, expected, desc}]) |

Extracting Helpers

Look for repeated setup (3+ occurrences) → extract to:

Go: testhelper/ or *_test.go helpers with t.Helper()
Python: conftest.py fixtures
TypeScript: test-utils.ts utilities

Mock Alignment

Standardize on one pattern per project:

Go: mockery --with-expecter --inpackage
Python: mocker.Mock(spec=Class)
TypeScript: vi.mocked() for type safety

Phase 6: Verify & Report

# Verify all tests pass
go test -v ./...
# or
pytest -v
# or
bun test

Output

TEST IMPROVEMENT COMPLETE
=========================
Mode: [selected mode]
Agent IDs: [list for resumption if needed]

Review:
- Issues found: X
- Issues addressed: Y

Refactoring:
- Tests combined: N → M table-driven
- Duplicates removed: X
- Helpers extracted: Y

Coverage:
- Before: XX%
- After: YY% (excluding non-business code)

Files modified:
- [list]

Execute test improvement now.

Agent Skills: Test Improvement

Install this agent skill to your local

Skill Files