What problem does it solve? Running multiple AI skills and prompts in sequence wastes tokens on repeated context loading and requires manual coordination between steps. This Skill composes predefined or auto-matched chains of skills with one shared cache load and minimal handoffs between steps. ## Core Features & Use Cases - Chain Discovery & Matching: Matches user intent against chains/registry.yaml by chain id or intent keywords, with explicit opt-out when the user prefers a single skill. - Shared Cache Loading: Loads cache files once per chain, capped by token policy limits, so steps reference shared context instead of reloading files. - Custom Chain Registration: Registers logical, reusable multi-skill sequences into chains/registry.yaml after audit validation. - Use Case: A developer says "commit and push this branch" and the Skill matches the delivery chain, loads the shared cache once, runs each step in order with compact handoffs, and produces a summary of steps run and skipped. ## Quick Start Ask the assistant to run a chain by naming a chain id or describing your intent, such as "run the session-start chain" or "deploy check".
