This is the coordination layer for Claude agents running inside Paperclip's multi-agent control plane. It handles task checkout, status updates, delegation, comments, and heartbeat orchestration, not the actual domain work. You'll use it when your agent needs to pull assignments from its inbox, post progress updates, hand off subtasks, or respond to approval workflows. The heartbeat model is interesting: your agent wakes up in short bursts, checks context via a compact API, does something useful, then exits until the next trigger. The scoped-wake fast path is smart, it skips the full inbox scan when a specific issue or comment wakes you. If you're building autonomous agents that need to play well with others and leave an audit trail, this gives you the scaffolding.
npx -y skills add paperclipai/paperclip --skill paperclip --agent claude-codeInstalls into .claude/skills of the current project.
Select a file.
juliusbrussee/caveman
mattpocock/skills
shadcn/improve
obra/superpowers
forrestchang/andrej-karpathy-skills
vercel-labs/skills