What problem does it solve? Agents running on different machines (laptop, DevBox, Azure VM) cannot easily share work or pass results without manual copy-paste handoffs. This Skill defines a git-based task queuing pattern so agents on separate machines can assign, execute, and return results automatically. ## Core Features & Use Cases - Git-Based Task Queue: Tasks are YAML files committed to .squad/cross-machine/tasks/ and results written to .squad/cross-machine/results/, giving an immutable audit trail through git history. - GitHub Issues Fallback: Urgent or ad-hoc tasks can be submitted as issues with squad:machine-{name} labels, which the target machine's agent detects and executes. - Security Validation Pipeline: Every task passes schema validation, command whitelisting, resource limits (timeout, memory, CPU), and execution isolation before running. - Use Case: A laptop agent needs a GPU workload (e.g., voice cloning) run on a DevBox. It writes a task YAML file, commits and pushes; the DevBox agent picks it up on its next 5-10 minute watch cycle, executes it, and pushes back a result file with exit code, stdout, and artifacts. ## Quick Start Create a YAML task file in .squad/cross-machine/tasks/ targeting the desired machine, then commit and push it so the remote agent executes it automatically.
