What problem does it solve? Adding support for a new AI agent runtime in Paperclip requires coordinated changes across three separate registries (server, UI, CLI), each with its own module interface, and mistakes in any one surface break agent execution, transcript rendering, or terminal output. ## Core Features & Use Cases - Adapter Package Scaffolding: Defines the required package layout under packages/adapters/<name> with the four-export package.json convention (root, server, ui, cli). - Server Execution Contract: Specifies the AdapterExecutionContext to AdapterExecutionResult flow, session codec design, environment variable injection, output parsing, and unknown-session retry logic. - UI and CLI Integration: Covers transcript parsing for the run viewer, config form field components, and colored terminal formatting for run watch mode. - Use Case: When adding support for a new CLI agent like a custom LLM tool, follow this Skill to implement execute.ts, parse.ts, config builders, and register the adapter in all three registries with correct session persistence. ## Quick Start Use the create-agent-adapter skill to scaffold a new Paperclip adapter package for my custom CLI agent and register it in the server, UI, and CLI registries.