Distributed Inference

What to check before installing/running this skill: - Review the upstream package and source (https://github.com/geeks-accelerator/ollama-herd and the PyPI project) to confirm the code matches the documentation and contains no unexpected network calls or telemetry. - Expect the coordinator and node agents to collect and store system metrics and per-request traces in ~/.fleet-manager (latency.sqlite and JSONL logs); if that data is sensitive, review/rotate/secure those files. - The system uses mDNS (_fleet-manager._tcp.local.) for discovery and may broadcast/listen on the local network; if you are on an untrusted network, restrict/mask that behavior or use explicit --router-url instead. - The SKILL.md shows installing via pip and uses python3 in examples; ensure you have the intended Python/pip version and do not install packages as root without verifying the package. - Because there is no registry install spec, the skill itself won't auto-install code, but following its instructions will install a third‑party PyPI package — treat that as a separate action and audit it. - If you need higher assurance, request the exact PyPI version hash or a signed release and compare it to the repository before installing. These checks will reduce risk and confirm the skill is doing what its documentation claims.