SKILL: test-generator

Purpose

Generate deterministic unit/integration tests that protect behavior and enable safe refactors.

When to Use

• A feature needs coverage.
• A bug fix needs a regression test.
• You want characterization tests before refactoring.

Inputs

• target (required, string): function/module/endpoint/contract behavior to test.
• test_type (optional, enum: unit|integration|e2e).
• framework (optional, string): e.g., pytest, jest, vitest, foundry, hardhat.
• constraints (optional, string[]): no network, deterministic time, limited mocks, etc.

Steps

1. Identify observable behaviors and edge cases.
2. Decide test boundaries:
• unit: pure logic
• integration: DB/service boundary
3. Create fixtures:
• deterministic data
• stable clocks/UUIDs (mock only when required)
4. Write tests:
• happy path
• failure paths
• security/validation checks (where applicable)
5. Run tests and iterate until they are stable.

Validation

• Tests are deterministic (no flaky time/network dependencies).
• Tests fail before the fix (for regressions) and pass after.
• Coverage targets the acceptance criteria, not implementation details.

Output

• Test files added/changed
• How to run tests
• Notes on fixtures/mocks used

Safety Rules

• Do not add tests that require real credentials or real external services by default.
• Avoid snapshot tests for highly variable outputs unless stabilized.

Example

Bug: “User creation accepts invalid email.”

Output: a failing test asserting 400 on invalid input, then passes after fix.