Update-Scan Skill

Goal

Keep the system current and self-evolving: detect what is new or better — in the local kit, in connected skill registries, and in the wider ecosystem (tools, MCP servers, models) — then recommend or auto-adopt the best option per scenario, always within the cost and permission bounds set by GABBE_AUTONOMY. This is the discovery engine behind the kit's self-evolution. It never adopts blindly: every candidate is validated and gated before it changes behavior.

Run it at preflight (Step 5 of preflight.skill) and on demand.

Sources to scan

1. Local kit — new/changed *.skill.md, guides, templates, personas since the last preflight snapshot (count + names).
2. Skill registries — configured universal SKILL.md registries (Agent Garden / google-skills, skills.sh, Agensi, agent-skills-hub, awesome-agent-skills …). Import path + validation is owned by skills-registry.skill.
3. Ecosystem — newer best-practice tools, MCP servers, and models, discovered via core/research.skill (respect the Research Policy source tiers). Note SOTA/expensive options explicitly.