The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes shell commands (find, cat) by interpolating parameters directly into a shell environment without input sanitization or path validation.
[DATA_EXFILTRATION]: The cat {{FILE_PATH}} command allows the agent to read any file on the system that it has permissions for. Since the file_path is user-provided and unrestricted, this could lead to the exposure of sensitive files such as SSH keys, AWS credentials, or environment variables.
[COMMAND_EXECUTION]: The use of heredoc redirection (cat > {{FILE_PATH}} << 'EOF') allows the skill to overwrite arbitrary files. This capability could be used to modify critical system configurations, shell profiles (e.g., .bashrc) for persistence, or other sensitive files if an attacker provides a malicious path.
[PROMPT_INJECTION]: The skill presents an indirect prompt injection surface as it processes content from external files and untrusted user input without sanitization or protective boundary markers.
Ingestion points: The file_path and new_content parameters, as well as the content of the file displayed via cat.
Boundary markers: No delimiters or instructions to ignore embedded commands are present in the procedure.
Capability inventory: The skill has the ability to read and write arbitrary files on the system using shell commands.
Sanitization: No validation or filtering of file paths or content is performed before execution.

update-text-file