The Agent Skills Directory

[COMMAND_EXECUTION]: The tf_to_ncl command in scripts/commands.py allows writing output to arbitrary file paths provided via the output argument. Without path sanitization, this could be used to overwrite critical system files or shell profiles to achieve persistence.
[DATA_EXFILTRATION]: The tf_to_ncl command in scripts/commands.py reads the content of any file specified in the tf_file argument. This could allow an attacker to read sensitive data, such as .aws/credentials or .ssh/id_rsa, and expose it by writing it to an accessible output file.
[COMMAND_EXECUTION]: The skill uses subprocess.run to call terraform and tfsec in scripts/commands.py. The path argument is passed directly to these binaries without validation, which may permit argument injection to manipulate tool behavior.
[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection through the ingestion of untrusted infrastructure files.
Ingestion points: File content read in tf_to_ncl and directories processed in tfsec_scan, tf_validate, and tf_plan (scripts/commands.py).
Boundary markers: Absent; there are no instructions to the agent to disregard instructions embedded within the Terraform files.
Capability inventory: The skill has file writing capabilities and the ability to execute external CLI tools via subprocess.run (scripts/commands.py).
Sanitization: Absent; the skill does not perform content or path validation on inputs.

tf-ncl