What problem does it solve? Notes in a knowledge vault often mix first-hand observations, cited findings, and received wisdom without any record of where each claim came from, making it impossible to audit trustworthiness later. This Skill classifies every factual claim in a single note against a five-value epistemic vocabulary and writes the results into the note's frontmatter after explicit user approval. ## Core Features & Use Cases - Per-claim epistemic classification: Extracts declarative factual claims (up to 25 per note) and labels each as first-hand, cited, inferred, received, or unmarked, with confidence and reasoning. - Approval-gated frontmatter back-fill: Presents the full claims analysis for review, then patches epistemic_status: and sources: frontmatter only after explicit approval, preserving the note body verbatim. - Aggregate status with 60% majority rule: Derives a note-level epistemic status, falling back to mixed with a single HTML comment pointer when no status dominates. - Use Case: After drafting a research note, run the audit to discover which claims lack sources, approve the classification, and have the note's provenance frontmatter written automatically before promoting it to evergreen. ## Quick Start Ask the assistant to classify the claims and back-fill provenance sources for a specific note by giving its name or path.
