Platform detection: see references/platform-detection.md and references/resolve-platform.sh

Review PR

Overview

/review-pr reviews a pull request as a claim: this PR says it implements task 65 — does it, and is the evidence there?

It is the missing third member of a family. /review-task and /review-story review a document before implementation. /review-code reviews a diff in isolation. Neither asks whether the change on the branch actually delivers what the work item promised, or whether the paper trail the pipeline left behind it is complete and honest.

Two read-only lenses run over the same scoped diff:

• Code — references/code-review-prompt.md, passed verbatim. The same reviewer /review-code, /qa-story and /qa-task dispatch.
• Conformance — references/pr-conformance-prompt.md. New here: coverage, scope, trail, consistency.

A PR can be flawless code that implements the wrong thing, or correct work whose gate never reached PASS. Neither lens sees the other's failures, which is why both run.

The skill is advisory. It writes a co-located report and optionally posts one summary PR comment. It never submits a formal approve/request-changes review, never writes a gate .yml (only qa-* skills do that), and never edits code.