HostLink

This skill legitimately provides host command execution — which is powerful and risky. Before installing or enabling it: - Treat HOSTLINK_TOKEN as a high-value secret. Do NOT store it in global config files unless you understand the risk; prefer ephemeral or workspace-scoped secrets and avoid committing it to disk or git. - The registry metadata is inconsistent: it does not declare the required HOSTLINK_TOKEN or config path it expects. Ask the publisher to correct metadata or do not trust automatic installation. - Vet the hostlinkd repository/binary (https://github.com/jebadiahgreenwood/hostlink) before installing on your host; build from source if you must, and review the code for privilege escalation/backdoors. - Run hostlinkd as a least-privileged user where possible; do not run it as root unless you accept the risk. - Limit socket access (unix_mode, group membership) and avoid enabling TCP unless protected by WireGuard and strict network controls. - If you enable this skill for an agent, restrict autonomous invocation or remove the token from global agent configs; consider requiring explicit human confirmation for any host-executed command. - Rotate the auth token after any change, and monitor hostlinkd logs for unexpected commands. Given the metadata/instruction mismatch and the ability to run arbitrary host commands, proceed only if you fully trust the daemon, the repo, and the people who will control the agent.