Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Run a red-green-refactor TDD workflow — generate failing tests first, implement to green, then check coverage gaps. Usage: /tdd <generate|coverage|validate> [target]
.claude/skills/alirezarezvani-tdd/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-04 | ✗→✓ | ▲ Improved | -11% | 0% |
| case-05 | ✗→✓ | ▲ Improved | -16% | 0% |
| case-06 | ✗→✓ | ▲ Improved | -28% | 0% |
| case-07 | ✗→✓ | ▲ Improved | 33% | 0% |
| case-09 | ✗→✓ | ▲ Improved | 8% | 0% |
Drive a test-first workflow for $ARGUMENTS using the TDD Guide skill. The first word of $ARGUMENTS selects the mode (generate, coverage, or validate); the rest is the target file or directory. If $ARGUMENTS is empty, ask which mode and target.
> Note on tooling: the tdd-guide scripts are Python library modules, not CLI tools — import them; do not invoke them as commands. Runnable patterns below.
/tdd generate <file-or-dir> — write failing tests FIRSTengineering-team/skills/tdd-guide/SKILL.md and engineering-team/skills/tdd-guide/references/tdd-best-practices.md for the red-green-refactor discipline and test-case taxonomy (happy path, edge cases, error cases)engineering-team/skills/tdd-guide/references/framework-guide.md for Jest/Vitest/pytest/JUnit conventionsbashcd engineering-team/skills/tdd-guide/scripts && python3 -c " from test_generator import TestGenerator, TestFramework g = TestGenerator(framework=TestFramework.PYTEST, language='python') cases = g.generate_from_requirements({'acceptance_criteria': [ {'id': 'AC1', 'description': 'validates email format'}, {'id': 'AC2', 'description': 'rejects duplicate emails'}]}) print(g.generate_test_file('registration', cases)) "
/tdd coverage <coverage-report> — analyze gaps against a thresholdpytest --cov --cov-report=lcov, vitest run --coverage, jest --coverage)bashcd engineering-team/skills/tdd-guide/scripts && python3 -c " from coverage_analyzer import CoverageAnalyzer a = CoverageAnalyzer() a.parse_coverage_report(open('<path-to-lcov-or-json>').read(), 'lcov') # or 'json' / 'xml' print(a.calculate_summary()) for gap in a.identify_gaps(threshold=80.0): print(gap) "
(Smoke-test input available at engineering-team/skills/tdd-guide/assets/sample_coverage_report.lcov.)
/tdd generate — coverage gaps are filled with tests, not excuses/tdd validate <test-file> — review test qualityRead the test file and check it against engineering-team/skills/tdd-guide/references/tdd-best-practices.md:
Report failures with concrete rewrite suggestions.
For wiring coverage thresholds into CI, follow engineering-team/skills/tdd-guide/references/ci-integration.md.
engineering-team/skills/tdd-guide/SKILL.md (+ HOW_TO_USE.md)engineering-team/skills/tdd-guide/references/tdd-best-practices.mdengineering-team/skills/tdd-guide/references/framework-guide.mdengineering-team/skills/tdd-guide/references/ci-integration.mdengineering-team/skills/tdd-guide/scripts/ (test_generator, coverage_analyzer, tdd_workflow, fixture_generator, metrics_calculator — import-only)engineering-team/skills/tdd-guide/assets/| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 2,293 | 8,276 | +261% | 1 | 1 | 0% | 356 | 1,460 | +310% | 0 | 0 | — |
case-02 | fail→fail | 9,104 | 6,509 | -29% | 1 | 1 | 0% | 1,526 | 1,455 | -5% | 0 | 0 | — |
case-03 | pass→pass | 10,589 | 9,907 | -6% | 1 | 1 | 0% | 2,077 | 3,015 | +45% | 0 | 0 | — |
case-04 | fail→pass | 9,177 | 2,942 | -68% | 1 | 1 | 0% | 1,726 | 1,534 | -11% | 0 | 0 | — |
case-05 | fail→pass | 8,225 | 2,362 | -71% | 1 | 1 | 0% | 1,675 | 1,403 | -16% | 0 | 0 | — |
case-06 | fail→pass | 12,909 | 4,105 | -68% | 1 | 1 | 0% | 2,575 | 1,855 | -28% | 0 | 0 | — |
case-07 | fail→pass | 7,278 | 3,831 | -47% | 1 | 1 | 0% | 1,268 | 1,684 | +33% | 0 | 0 | — |
case-08 | pass→pass | 3,309 | 1,781 | -46% | 1 | 1 | 0% | 607 | 1,269 | +109% | 0 | 0 | — |
case-09 | fail→pass | 8,735 | 3,746 | -57% | 1 | 1 | 0% | 1,498 | 1,614 | +8% | 0 | 0 | — |
case-10 | pass→pass | 8,140 | 2,976 | -63% | 1 | 1 | 0% | 1,428 | 1,484 | +4% | 0 | 0 | — |
case-11 | pass→pass | 7,543 | 4,162 | -45% | 1 | 1 | 0% | 1,092 | 1,626 | +49% | 0 | 0 | — |
case-12 | fail→pass | 7,705 | 4,126 | -46% | 1 | 1 | 0% | 1,516 | 1,683 | +11% | 0 | 0 | — |
case-13 | pass→pass | 4,302 | 3,307 | -23% | 1 | 1 | 0% | 797 | 1,491 | +87% | 0 | 0 | — |
case-14 | fail→pass | 8,498 | 1,739 | -80% | 1 | 1 | 0% | 1,695 | 1,310 | -23% | 0 | 0 | — |
case-15 | fail→pass | 7,540 | 1,966 | -74% | 1 | 1 | 0% | 1,286 | 1,348 | +5% | 0 | 0 | — |
case-16 | fail→pass | 7,562 | 2,985 | -61% | 1 | 1 | 0% | 1,453 | 1,572 | +8% | 0 | 0 | — |
case-17 | pass→pass | 10,158 | 4,876 | -52% | 1 | 1 | 0% | 1,814 | 1,887 | +4% | 0 | 0 | — |
case-18 | pass→pass | 10,026 | 3,449 | -66% | 1 | 1 | 0% | 1,757 | 1,616 | -8% | 0 | 0 | — |
case-19 | pass→pass | 5,660 | 5,492 | -3% | 1 | 1 | 0% | 1,055 | 1,865 | +77% | 0 | 0 | — |
case-20 | pass→pass | 12,000 | 8,973 | -25% | 1 | 1 | 0% | 2,224 | 2,519 | +13% | 0 | 0 | — |
case-21 | pass→pass | 12,476 | 8,480 | -32% | 1 | 1 | 0% | 2,462 | 2,419 | -2% | 0 | 0 | — |
case-22 | pass→pass | 14,536 | 12,771 | -12% | 1 | 1 | 0% | 2,715 | 3,210 | +18% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 22 cases were attempted, and 20 counted toward the lift figure. The other 2 produced results that are not comparable between the two arms, so they are excluded from the headline rather than averaged into it. The headline lift of +41 percentage points is the difference between those two pass rates over the 20 comparable cases.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
Other measured skills in the registry, with their headline benchmark lift.