ws-tdd-red
Phase 3 -- implement failing tests, validating assertion quality and depth.
- License 许可
- MIT
- 最近更新
- 2026-10-06
- Stars
- ★ 4
- Forks
- 0
01 前言 Front Matter
- name
- ws-tdd-red
- description
- Phase 3 -- implement failing tests, validating assertion quality and depth.
03 SKILL.md 正文
ws.tdd.red
Load shared context: ws.tdd._meta/SKILL.md — the sibling skill in the same skills directory this skill loaded from (resolves under any clone name, install location, or skills-root override). Also read ws.tdd._meta/visibility-patterns.md and ws.tdd._meta/traceability.md.
Purpose
Turn outlines into executable test code. Not "does test run" but "does test assert right thing at right depth?" Cognitive mode: Validation of validations. Scrutinize every assertion -- checking what matters? deeply enough? subtly wrong impl still pass?
Prerequisites: Test outlines from /ws.tdd.outline (Phase 2), reviewed and approved by human.
Process
1. Select Outlines
Ask: "Which outlines to implement?" (or confirm if human specified). Max 3 per turn.
BDD awareness: .feature files have @todo scenarios for this feature -> "Note: N BDD scenarios still @todo -- should route through /ws.bdd.outline -> /ws.bdd.red." Not blocker, surfaces gap.
2. Implement
Remove todo marker and skip, replace with real assertions. Outline comments must survive -- "why/what" above "how." Comments follow AAA structure, explain intent not mechanics. Comment redundant with code -> outline at wrong abstraction -- raise with human, don't silently drop.
# Goal: Verify fail-fast behavior when input file doesn't exist --
# missing files must be caught at construction time, not at first use
# Boundaries: Only tests missing file; malformed content is separate.
# Does not test the error message content.
def test_output_mapper_throws_on_missing_file():
# Arrange: use a path that cannot exist
bad_path = "/nonexistent/path/data.xml"
# Act + Assert: construction itself must fail, not a later method call
with pytest.raises(FileNotFoundError, match=bad_path):
OutputMapper(bad_path)
// Goal: Verify fail-fast behavior when input file doesn't exist --
// missing files must be caught at construction time, not at first use
// Boundaries: Only tests missing file; malformed content is separate.
// Does not test the error message content.
TEST_CASE("OutputMapper throws on missing file", "[error]") {
// Arrange: use a path that cannot exist
auto bad_path = "/nonexistent/path/data.xml";
// Act + Assert: construction itself must fail, not a later method call
REQUIRE_THROWS_AS(OutputMapper(bad_path), std::runtime_error);
REQUIRE_THROWS_WITH(OutputMapper(bad_path), Catch::Contains(bad_path));
}
3. Run Test Suite
Run entire suite. New tests should fail -- passing immediately -> investigate: vacuously passing (success theater)? existing code handles case (acceptable but steps too large)? Existing tests must still pass.
4. Review Failures
Each: right reason (missing impl, not test bug)? / informative message? / assertion depth matches outline intent?
5. Coverage Report (when SDD context available)
"X of Y acceptance criteria have executable tests. Outline only: [list]. Uncovered: [list]."
Output
Executable, failing test code. Suite has been run.
Behaviors
- Every assertion = design decision: "
assert len(result) == 3enough, or also checkresult[0].name == 'expected'?" - Exact values, not defensive ranges.
== 1not>= 1.== "success"notis not None. Weak assertions worse than no test. See "Assertion Precision" inws.tdd._meta. - "Verify content" = verify content, not presence/count. Under-specified outline: note it, don't silently decide.
- Test independence: no test depends on another's execution/state. Extract shared setup: fixtures/helpers for common setup, vary only what differs.
Anti-Patterns
Success theater (assert True / unfailable assertions) / writing production code (Phase 4) / skipping new tests (wrote it = runs) / weakening assertions ("check content" -> "check not empty") / exceeding 3 per turn / catching exceptions to prevent failure.
Assertion Precision Checkpoint
After each batch, before gate: "Would subtly wrong impl still pass?" > 0/is not None where exact value known -> == exact_value / presence not content -> add content / type not value -> add value / one field ignoring related -> add missing. Fix before gate.
Gate
Present tests + failure output. Wait for confirmation: (1) fail for right reasons (2) assertion depth correct (3) ready for /ws.tdd.green.
After Green
Return for next batch (max 3/turn). All outlines done -> back to /ws.tdd.outline. See "Batch Processing Flow" in ws.tdd._meta.
04 同仓库更多 Skills
ws-tdd-review
Audit existing test code for quality, depth, and adherence to TDD principles.
ws-tdd-outline
Phase 2 -- tree-level test design, examining intent and approach per test.
ws-tdd-idea
Phase 1 -- forest-level test case discovery, mapping the full problem space.
ws-tdd-green
Phase 4 -- write minimal production code to make failing tests pass.
ws-tdd-meta
TDD shared context -- phased protocol, quality standards, and standing instructions for all TDD skills.
ws-sweep
Completeness sweep -- exhaustively verify nothing was missed after a change.