skill-freshness

Detect silent staleness by auditing on-disk file mtimes against cadence-derived freshness thresholds defined in aeon.yml. This prevents autonomous skill chains from consuming outdated upstream files without visible failures by surfacing consumer-ready warnings.

Track upstream dependency freshness by auditing explicit consume edges and parsing implicit references from SKILL.md files. This mechanism scores dependency chains for scheduled jobs, comparing local file mtimes against producer cadence thresholds to generate a decision-ready report.

Audit file mtime freshness by running a local-only discovery process that checks enabled consumer skills scheduled in aeon.yml. It performs mtime-based freshness checks with cadence-derived thresholds to ensure agent reliability and outputs deduped notifications.

Yes, freshness auditing works with implicit dependencies by using best-effort references parsed directly from each consumer SKILL.md. It combines these parsed implicit references with deterministic consume edges to comprehensively track explicit and implicit dependency chains.

Stop repeated staleness alerts through deduped reporting and notifications that write daily reports and memory state. This mechanism suppresses repeated alerts when the flagged set of outdated upstream articles or outputs hasn't changed, reducing notification fatigue.

Local-only staleness detection is limited to on-disk file discovery and mtime checks within the local environment. It cannot validate freshness for distributed memory artifacts or remote outputs unless they are locally synced and discoverable by the file mtime auditing process.