Test Generator

Automatically create or update tests for files changed in the current branch,
iterating until statement coverage reaches the target threshold (default 80%).

Workflow

1. Detect changed files — run scripts/detect-changes.sh [base-branch]
2. Discover project conventions — read existing tests, test config, and setup files
3. Write / update tests — for each uncovered file, create or extend its test file
4. Run coverage check — run scripts/check-coverage.sh <threshold> [files...]
5. Iterate — if any file is below threshold, read the coverage report, add missing tests, repeat from step 4 (max 3 iterations)
6. Report — summarize final per-file coverage to the user

Step 1 — Detect Changed Files

bash <skill-path>/scripts/detect-changes.sh main

• Pass the base branch as argument (defaults to upstream or main).
• Output: one source file path per line.
• Filters out: test files, config, styles, assets, lock files, __mocks__/, test/setup.

If no files are output, inform the user there are no testable changes.

Step 2 — Discover Project Conventions

Before writing any test, read these to match existing style:

1. Test config — vitest.config.*, vite.config.* (test section), Cargo.toml
2. Test setup — files referenced by setupFiles in config
3. Existing test for the file — <name>.test.ts, <name>.spec.ts, or __tests__/<name>.ts
4. Neighboring tests — 1-2 test files in the same directory for style reference
5. Project rules — AGENTS.md, CLAUDE.md, .eslintrc*, eslint.config.*

Key things to extract:

• Test framework and assertion style (e.g. expect(), assert)
• Import conventions (import { describe } from 'vitest' vs globals)
• Mock patterns (manual mocks, vi.mock())
• File naming: *.test.ts vs *.spec.ts
• Any JSDoc / lint requirements on test files

Step 3 — Write / Update Tests

For each changed file without sufficient coverage:

If no test file exists → create one following the project naming convention.

If a test file exists → add new test cases; do not rewrite existing passing tests.

Test quality guidelines

• Test behavior, not implementation details.
• Cover: happy path, edge cases, error paths, boundary values.
• Use descriptive describe/it block names that explain the expected behavior.
• For React components: prefer @testing-library/react queries (getByRole, getByText) over querySelector.
• For functions: test return values and side effects, not internal calls.
• Mock external dependencies (API, file system, DB) but not the unit under test.
• If the project uses fast-check, consider property tests for pure utility functions (name: *.property.test.ts).

Step 4 — Run Coverage Check

bash <skill-path>/scripts/check-coverage.sh 80 src/services/foo.ts src/utils/bar.ts

• First argument: threshold percentage.
• Remaining arguments: filter output to only these files.
• Output lines: PASS 92.3% src/services/foo.ts or FAIL 65.0% src/utils/bar.ts.

If the script fails to find coverage output, check that the coverage provider is installed
(@vitest/coverage-v8, @vitest/coverage-istanbul, etc.) and install if missing.

Step 5 — Iterate

For each FAIL file:

1. Read the coverage JSON (coverage/coverage-final.json for vitest) to identify uncovered statements.
2. Add targeted test cases for uncovered branches/statements.
3. Re-run coverage check.
4. Repeat up to 3 iterations total. If still below threshold after 3 iterations, report remaining gaps to the user with suggestions.

Step 6 — Report

Summarize results:

## Test Coverage Report

| File | Coverage | Status |
|------|----------|--------|
| src/services/foo.ts | 92.3% | PASS |
| src/utils/bar.ts | 81.0% | PASS |
| src/components/Baz.tsx | 73.5% | FAIL |

Target: 80% | Passed: 2/3

For FAIL files, briefly explain what remains uncovered and suggest next steps.

Scripts

• scripts/detect-changes.sh [base-branch] — list changed source files needing tests
• scripts/check-coverage.sh <threshold> [files...] — run tests with coverage and report per-file results