Orchestrate multi-step project execution across fleet members with review loops.
npx skills add https://github.com/Apra-Labs/apra-fleet --skill pm-apra-labs
Please help me install this Agent Skill. Skill: pm Source: https://github.com/Apra-Labs/apra-fleet/tree/main/skills/pm Command: npx skills add https://github.com/Apra-Labs/apra-fleet --skill pm-apra-labs
It solves the problem of turning a goal into a coordinated, multi-step build-and-review workflow across multiple fleet members, with clear checkpoints and recovery.
Use the pm skill to create a new sprint and run it end-to-end by issuing: /pm init <project> then /pm plan <requirement> and /pm start <member>.
High-intent search queries and answers about installing and using this skill.
Multi-agent project execution is coordinated by planning phased tasks, dispatching doer/reviewer pairs, tracking persistent task state, and managing review loops to drive sprints to completion. The PM skill orchestrates this workflow without writing code directly.
A doer/reviewer review loop enforces quality gates by assigning one agent to execute code and another to review it. This cycle includes verification checkpoints, backlog handling for deferred findings, and final PR raising to guarantee CI verification without merging.
To initialize a sprint end-to-end, issue /pm init <project> to set up the project, then /pm plan <requirement> to generate a phased plan, and finally /pm start <member> to dispatch execution tasks to fleet members.
Yes, agent orchestration requires a working fleet skill dependency to dispatch and manage fleet members. The PM skill relies on this fleet infrastructure and uses Beads for lifecycle integrity to maintain persistent cross-project task graphs.
Yes, deployment runbook execution is supported by preparing a deploy.md file and executing each step on the target fleet member. This happens after the doer/reviewer review loop and PR verification are complete, ensuring safe deployments.
Doer-only code execution is enforced to maintain strict separation of duties within the review loop. By preventing the PM agent from writing code directly, the system guarantees that all code changes pass through verification checkpoints and reviewer approval before PRs are raised.
supportersimulatorDispatch remote tasks to fleet nodes with per-task tracking.
jimloopCoordinate Git commits, repository pushes, and cloud deployment workflows.
gioeDetect duplicate tasks in SQLite databases using character-level and token-level similarity scoring.
nateaCoordinate quality engineering across CI/CD pipeline phases with skill selection and agent coordination.
