documentation-claim-review

What problem does it solve? Public-facing documentation such as READMEs, release notes, and setup guides often contains stale, overstated, or unsupported claims. This Skill fact-checks each claim against designated repository evidence so teams can publish accurate docs without silently rewriting unsupported statements. ## Core Features & Use Cases - Atomic Claim Splitting: Breaks compound prose into individual claims and classifies them by type (behavior, compatibility, performance, security, release status, and more). - Evidence-Ranked Verification: Applies a strict evidence hierarchy from revision-bound runtime proof down to comments, rejecting stale, development-only, or overly narrow evidence. - Labeled Findings Report: Assigns each claim a label (verified, partially verified, unsupported, stale-suspected, contradicted, needs-human) with confidence levels and next-step evidence requirements. - Use Case: Before a release, ask the Skill to verify that the release notes match the actual implementation; it returns a claim-by-claim assessment with publication blockers and evidence gaps instead of guessing. ## Quick Start Ask the assistant to fact-check the README's claims about supported platforms against the current repository implementation and report any unsupported statements.