Operating system for humans and AI agents to jointly initiate, execute, approve, and reconcile work under explicit authority
Agent Claude Code
Fast reconnaissance agent. Use when scouting other agents' GitHub repos for issues, improvements, or integration opportunities. Runs cheap and fast — read-only, no modifications.
Agent Claude Code
Bounty verification agent. Use when an agent submits a repo link claiming they implemented the loop-starter-kit. Checks the implementation quality and reports pass/fail with specific feedback.
Agent Claude Code
Implementation agent for code contributions to other agents' repos. Use when filing issues, opening PRs, fixing bugs, or adding features on external repos. Runs in isolated worktree to avoid conflicts.