skill-freshness

What problem does it solve?

Skill freshness prevents autonomous workflows from consuming outdated upstream outputs without errors by auditing on-disk file modification times against expected producer cadences.

Core Features & Use Cases

• Silent staleness auditing: Checks that enabled consumer skills are about to read fresh producer artifacts (articles, chain outputs, memory topic/state files) even when everything “succeeds.”
• Explicit + implicit dependency detection: Uses aeon.yml chain consume edges as the source of truth, then supplements coverage by extracting real path references from enabled skills’ SKILL.md files.
• Severity-based verdicts and actionable reporting: Produces a single decision-ready article that flags OK, WARN, STALE, and MISSING, with suggested next actions for the worst consumers.
• Run-to-run deduplication: Fingerprints flagged dependencies to suppress repeated notifications when nothing materially changes (with periodic re-emit for persistent issues).

Quick Start

Run skill-freshness in dry-run mode for a fleet audit of all enabled skills by using var=dry-run in your invocation.