The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The ssh_execute action enables the agent to run arbitrary shell commands on a remote server via the SSH protocol.
[CREDENTIALS_UNSAFE]: The skill's design requires passing sensitive authentication data, including plaintext passwords within connection URLs and SSH private keys in PEM format as action parameters.
[EXTERNAL_DOWNLOADS]: The skill instructions direct users to install and update components from the author's GitHub repository using npx skills add.
[COMMAND_EXECUTION]: The skill facilitates the execution of remote commands and file system modifications (e.g., ftp_delete) on external infrastructure.
[PROMPT_INJECTION]: The skill provides a surface for indirect prompt injection by ingesting unvalidated data from remote servers which is then processed by the agent.
Ingestion points: Remote file content retrieved via ftp_download and shell output from ssh_execute.
Boundary markers: The skill documentation does not define delimiters or isolation markers for the ingested data.
Capability inventory: The skill can perform remote command execution, file uploads, and file deletions.
Sanitization: No sanitization or validation of remote server responses is performed before the data enters the agent's context.

multi-protocol-bridge