What problem does it solve?
Helps transform messy, mixed-credibility information into structured knowledge units while preventing low-quality or potentially contaminated content from polluting long-term memory through risk scoring and quarantine/inheritance queues.
Core Features & Use Cases
- Gradient knowledge digestion: Scores each item on core claim, evidence, and application using configurable keyword gradients plus semantic signals, then outputs a digestion report.
- Freshness & review scheduling: Computes category-based freshness windows and generates a review (反刍) schedule to keep knowledge actionable over time.
- Contamination risk control: Produces a contamination risk score and separates knowledge into inheritance_queue (safe to keep) and quarantine_queue (requires review).
- Use case: You ingest research notes, blog posts, and forum discussions; this skill digests them into units with freshness days, decides which ones to inherit vs quarantine, and tells you when to review each unit.
Quick Start
Use the liuku-xianzei skill to digest an input file named info.json and produce digest_report.json with knowledge units, freshness windows, and review plans.