What problem does it solve?
Running multiple autonomous AI agents in parallel on a shared VPS often leads to uncoordinated file edits, double-claimed tasks, misaligned reviews, and wasted work from poor cross-agent communication and unclear task ownership.
Core Features & Use Cases
- Per-agent autonomy toggles: Assign YOLO mode for low-risk tasks like scaffolding and bulk refactors, and review-gated mode for high-stakes architectural changes, matching execution speed to risk level.
- Multi-model layered review: Leverage Claude for high-level reasoning, Codex for implementation consistency checks, and Gemini for cross-specification alignment to catch defect classes individual models miss.
- Beads + Agent Mail coordination: Use beads for global task DAG tracking and stable claim IDs, and Agent Mail for async communication of handoffs, review findings, and reservation announcements.
- Collision recovery workflows: Standardized steps to resolve double-claim conflicts and handle wedged agents without uncoordinated unilateral action.
- Use Case: A development team running a 3-agent swarm to implement 12 parallel feature beads can use this skill to assign tasks, avoid edit races on shared files, run cross-model reviews, and ensure the final assembled output passes full template QA.
Quick Start
Invoke this skill to coordinate a multi-agent VPS swarm for parallel feature development, assign task beads to specialized agents, resolve double-claim collisions, and enforce layered cross-model reviews end-to-end.
