Two lean sentences with concrete commands and a justified rationale for the repo-wide check; no padding and no concepts Claude already knows, so every token earns its place.

Gives the fully executable command `pnpm policy-check --path <relative-path>` with where to run it (repo root), the scope variant (`pnpm policy-check` without `--path`), and an explicit anti-pattern callout (do NOT cd into the package).

As a simple single-purpose skill, the single action (run policy-check) is unambiguous, with clear scope guidance and a rationale; it is a read-only compliance check, so no validation checkpoint is required and the cap-at-2 rule does not apply.

A self-contained skill under 50 lines with no external references needed (no bundle files exist); the body is well-organized and complete on its own, satisfying the simple-skill allowance.

Names the domain (policy checks) and a couple of actions (running checks for a specific package vs the entire repository, validating compliance), but does not list multiple distinct concrete actions — essentially one operation at two scopes.

Clearly answers both what ("Provides guidance on running policy checks...") and when, with an explicit "This skill should be used when the user asks to..." trigger clause.

Provides several natural phrasings a user would actually say — "run policy check", "check policy", "policy-check", and "validate package compliance" — giving good coverage of variations.

Targets a clear, distinct niche (package policy compliance) with specific trigger phrasings, making it unlikely to fire for the wrong skill.