← ClaudeAtlas

pattern-test-coveragelisted

Role-neutral catalogue of what makes a test set *complete* — the shared substance both the engineer (authoring tests in the TDD red phase) and the reviewer (gating the code) judge against. Every EARS criterion is discharged by the cheapest durable proof at its owning layer (backend / frontend / E2E); new code paths cover edge breadth. The spine is the deletable-code lens: complete only when deleting any single production branch makes a test fail. Activate when writing or reviewing tests.
MartinKChen/harness-claude-code · ★ 0 · AI & Automation · score 72
Install: claude install-skill MartinKChen/harness-claude-code
# pattern-test-coverage The single, role-neutral catalogue of **what counts as a complete test set** for a unit of work. "Complete" is the same fact whether you are *writing* the tests (the engineer, in the TDD red phase) or *judging* them (the reviewer, at the code gate) — a deletable branch is deletable from either chair, and an off-by-one boundary is missing whether you are authoring it or flagging its absence. This skill owns that shared substance so it lives in exactly one place. What this skill does **not** own — the role-specific framing layered on top of it: - **Authoring discipline** (write the test before the code, one behavior per RED, fakes at seams) — `principle-engineer-tdd`. - **Detection, severity, and reporting** (every gap is HIGH and blocks the gate, the `file:line` finding shape, the `# Code Review` comment template) — `pattern-reviewer-test-coverage`. Both of those skills reference this catalogue for the substance and add only their own verb. ## When to activate - An engineer is **authoring or extending tests** for a behavior (every TDD RED step, every fix that adds coverage). Walk the catalogue and close every gap that applies before calling the behavior green. - A reviewer is **judging the code gate** on any `type:backend` / `type:frontend` / `type:e2e` task or slice. Walk the catalogue to find gaps; `pattern-reviewer-test-coverage` then grades and reports them. - A user asks "are the tests enough", "did we cover the acceptance criteria", "what ed