What problem does it solve?
Poorly authored agent skills — vague descriptions, bloated bodies, dead file links, or committed secrets — silently degrade every agent run that loads them, and teams have no systematic way to catch these issues as their skills store grows.
Core Features & Use Cases
- Fresh-skill sweep: Checks recently changed skills against an 8-rule checklist covering description quality, name format, body size, single responsibility, link hygiene, secrets, instruction style, and duplication.
- High-leverage deep pass: On a weekly rotation, audits the store's most-used or most-referenced skills even when unchanged, ranked by usage telemetry, version churn, or cross-references.
- Report authoring and dedupe: Files one P3 recommendation report per non-compliant skill via the inbox report channel, editing existing reports on regressions instead of duplicating, with per-version dedupe tracked in a scratchpad.
- Use Case: A team maintains dozens of shared agent skills in PostHog; this scout runs on a schedule, detects that a skill bundles a dead reference link and a leaked API key, and files a copy-ready fix report routed to the skill's author.
Quick Start
Run the skills-store scout to audit recently changed skills in our PostHog skills store and file reports for any authoring violations found.
