What problem does it solve? Rule-based checks on a generated CLI's output can only catch what regexes encode, so semantic bugs like irrelevant search results, silently dropped sources, broken ranking, and malformed URLs ship undetected. This Skill dispatches a reviewer agent to judge sampled command output the way a human tester would. ## Core Features & Use Cases - Semantic relevance review: Judges whether query results actually match user intent beyond mechanical substring matching. - Format and URL inspection: Detects mojibake, raw HTML entities, and URLs pointing at index pages instead of canonical permalinks. - Aggregation and ranking validation: Verifies fan-out commands show all requested sources and that ranked output is plausibly ordered. - Structured result contract: Emits a parseable PASS, WARN, or SKIP block that parent skills consume during shipcheck. - Use Case: During a printing-press shipcheck at Phase 4.85, the parent skill invokes this reviewer against the printed CLI's live-check samples, and any plausibility findings are logged as non-blocking warnings for the user to triage before publishing. ## Quick Start Invoke this skill from the printing-press shipcheck by passing the printed CLI's working directory so it reviews the sampled live-check output and returns a structured findings block.
