What problem does it solve? Growing a media collection without a plan leads to padded libraries and duplicate purchases. This Skill audits a Plex server's actual contents, models the owner's taste from watched items, finds structural and canon gaps, and proposes additions where every pick is verified absent and backed by a named piece of evidence. ## Core Features & Use Cases - Library census and taste modeling: Uses Plex MCP tools like library_export, library_stats, and discover to inventory the collection and infer preferences from watched items, with horror volume explicitly excluded from taste weighting. - Warrant-based recommendations: Every proposed title carries a labeled warrant type (Canon, Bridge, Lineage, Coverage, or Event) and is verified absent via check_titles or search before it ships. - Batch curation with caching: Supports long-running curation passes with a 90-day canon-list cache in references/canon-lists and incremental append-only report writing to /sandbox/out. - Use Case: Ask "what am I missing from my movie library?" and receive a report showing zero documentaries, a missing Blade Runner (1982) behind its owned sequel, and a shortlist of defensible additions. ## Quick Start Ask the agent to audit my Plex library and propose ten defensible additions, writing the full report to /sandbox/out.
