skill-freshness

Audit dependency freshness by scanning all enabled skills to evaluate file age against per-class thresholds, classifying upstream references as OK, WARN, STALE, or MISSING to prevent downstream reads of stale data.

Stale data reads occur when upstream file dependencies like articles, outputs, and memory state files become outdated. Auditing mtimes against cadence-derived thresholds detects these issues before production runs manifest them.

Monitor upstream dependencies by evaluating explicit chains and implicit references across articles, outputs, topics, and state files. The audit computes age against per-class thresholds and writes a daily report with severity classifications.

Yes, the freshness audit generates a daily article and persists a fingerprinted state to prevent duplicate notifications. Optional operator alerts are triggered when issues or missing dependencies arise.

Yes, the audit applies to every skill defined in aeon.yml. It evaluates explicit dependency chains and implicit references across all enabled skills to compute age and assign severity.

Freshness monitoring relies on file mtimes and per-class thresholds derived from producer cadence. It cannot detect logically incorrect data, only age-based staleness, and classifies dependencies strictly as OK, WARN, STALE, or MISSING.