agentplane

# AgentPlane Use AgentPlane through its CLI instead of editing `.agentplane/` state directly. ## When to use - The repository already contains `AGENTS.md` or `.agentplane/`. - The user wants to initialize AgentPlane in a repository. - The user wants explicit planning, task IDs, verification evidence, or `branch_pr` worktrees. ## Startup 1. If the repository is not initialized, run `ap init` or `agentplane init`. 2. Select ready work with `ap task active`. 3. Request one bounded external-agent action with `ap task advance <task-id> --agent-json`. 4. When `action.kind=agent_episode`, perform only the supplied semantic objective inside the supplied authority and write the typed result to `exchange.result_path`. 5. Resume with the exact `exchange.resume_argv` and request a fresh packet after every state change. `exchange.return_invocation` is compatibility-only. 6. If the repository has a configured managed runner, use `ap task run <task-id>` instead of the external exchange loop. ## Rules - Treat `AGENTS.md`, `ap quickstart`, and `ap role <ROLE>` as the policy surface. - Use `ap task advance` or `ap task run` as the normal protocol; do not edit `.agentplane/` state directly. - Do not invoke work start, start-ready, verify, finish, integrate, cleanup, Git branch/worktree, commit, or PR lifecycle commands during a normal semantic episode. - Treat approval, human, hosted/external, stale-state, and effect-in-doubt stops as hard boundaries; return control instea...