The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/ssh-deploy.sh executes arbitrary shell commands on a remote server through the run_commands function. These commands are retrieved from the pre_deploy, post_deploy, and rollback fields in the config.yaml file and executed sequentially via a separate ssh-exec.sh utility. Although this is the intended functionality of a deployment tool, it grants significant control over the remote environment.
[INDIRECT_PROMPT_INJECTION]: The skill has an attack surface for indirect instructions because it processes data from an external configuration file without validation.
Ingestion points: config.yaml is read and parsed by the read_config function in scripts/ssh-deploy.sh.
Boundary markers: None. The skill does not use delimiters or instructions to ignore potentially malicious content within the configuration file.
Capability inventory: The skill can execute arbitrary remote commands (ssh-exec.sh) and upload local files to remote paths (ssh-upload.sh).
Sanitization: None. Values from the YAML file are directly interpolated into shell commands and executed. If an attacker influences the content of config.yaml (e.g., via a malicious pull request or external data source), they could execute unauthorized commands on the target server.

ssh-dev-suite/deploy