The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze Infrastructure as Code (IaC) files such as Terraform, Kubernetes, and CloudFormation from the local directory. This represents a surface for indirect prompt injection where malicious instructions embedded in configuration files could potentially influence the agent's interpretation of security reports or affect downstream CI/CD actions.
Ingestion points: The skill processes local directories and files using the -d and -f flags, as seen in the usage examples in SKILL.md.
Boundary markers: No specific delimiters or safety instructions are provided to the agent to distinguish between its instructions and the potentially untrusted code in the files being scanned.
Capability inventory: The skill executes commands to generate structured reports (SARIF, JSON, JUnit XML) which are subsequently used for automated decision-making or reported to the user.
Sanitization: The skill relies on the underlying checkov utility for data handling, with no additional sanitization logic described.
[DYNAMIC_EXECUTION]: The skill documents the capability to run custom Python checks using the --external-checks-dir flag. This allows the tool to dynamically load and execute Python scripts from a specified local directory, as demonstrated in references/custom-checks-and-ci.md where a template for a custom EC2 tag check is provided.
[EXTERNAL_DOWNLOADS]: The skill instructions include installing the checkov package via pip install and pinning versions for consistency in CI environments. This fetches the utility from the standard Python package registry.

checkov-policy