The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes several local shell commands to manage and verify SSH configurations. These include ssh -G for parsing configuration results, ssh-keyscan for gathering public host keys, and custom tools like auto-config, validate, and render. The instructions mandate that these commands be used within specific authorization boundaries and generally prioritize local validation over network activity.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data from devices.toml and trust/known_hosts. While this presents a potential injection surface, the skill includes robust mitigation strategies, such as strict schema validation defined in references/device-schema.md, a requirement to fail on unknown fields or duplicate names, and an instruction to render and review outputs in temporary directories before application.
[SAFE]: The skill demonstrates best practices for secret management by explicitly prohibiting the inclusion of private keys, passwords, tokens, or agent sockets in the inventory or Git history. It also enforces a 'read-only' default state, requiring explicit user permission for any changes to live configurations or remote connections.

ssh-fleet