What problem does it solve? Page-level consistency checks miss subtle factual conflicts: two wiki entries that broadly agree but disagree on one benchmark number, or a stale count embedded in an otherwise-current entry. This Skill makes every factual assertion explicit, classifies it by evidence type, and flags contradictions at the sentence level. ## Core Features & Use Cases - Claim Extraction & Classification: Pulls every factual assertion from wiki entries and labels it as direct-quote, sourced, synthesis, or inference, with confidence scores and subject/predicate/object structure. - Contradiction Detection: Scans for cross-entry conflicts and intra-entry drift (TL;DR overstating the body, mismatched tags, stale counts), typed as direct_conflict, stale_data, scope_difference, framing_difference, or intra_entry_drift. - Structured Outputs: Writes a growing claims index to _inbox/claims-index.json and a severity-grouped report to _inbox/reports/claims-report-<date>.md; integrates with /wiki-cycle, /wiki-update, and /wiki-lint. - Use Case: After ingesting a new entry about a project's benchmark results, run /wiki-claims --compare <entry> to check its claims against the existing index and surface any conflicting numbers before filing. ## Quick Start Ask the AI to run /wiki-claims across the wiki folder to extract all factual claims and report any contradictions between entries.
