The body is dense with non-obvious reference material (a 7-category taxonomy, executable worked example, output template, anti-patterns table, limitations) and avoids explaining concepts Claude already knows; every section earns its place rather than padding.

Provides fully executable pytest code with real imports and concrete assertions, a copy-paste output-format template, and an anti-patterns table with explicit Fix columns mapping each failure to a concrete remediation.

The generation task is unambiguous and sequenced via numbered categories (1-7) plus a 'Recommended next step' feedback loop (run tests -> diagnose any failure as a real gap -> commit on pass), satisfying the validation checkpoint requirement for this batch operation.

The body is a concise overview with a single, clearly signaled one-level-deep reference to references/negative-path-catalog.md, which exists and holds the detailed per-pattern tables; navigation is easy and content is appropriately split.

Lists multiple concrete actions - 'produces companions exercising input validation rejection, missing required fields, type mismatches, authorization failures, rate-limit errors, and adversarial payloads' and 'Emits cases as parameterized tests in the project's runner format' - matching the score-3 anchor of several specific concrete actions.

Explicitly answers both what (generates negative/error-path companions mirroring happy-path tests) and when via a clear 'Use when a feature has happy-path coverage but the rejection / error / unauthorized paths are untested' trigger clause.

Natural terms a developer would actually say appear throughout: 'happy-path coverage', 'rejection / error / unauthorized paths are untested', 'negative / error-path test cases', giving good coverage rather than jargon.

A clear niche - generating negative companions to existing happy-path tests, paired with malicious-payload-bank - with distinct triggers unlikely to fire for unrelated skills.