The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/generate_network_policy.py includes logic to generate shell scripts and explicitly sets executable permissions using file_path.chmod(0o755).
Evidence: scripts/generate_network_policy.py uses chmod(0o755) on files generated from the template variable.
[REMOTE_CODE_EXECUTION]: The skill implements a dynamic code generation pattern where the generate_network_policy.py utility assembles executable .sh files using content provided via the --content command-line argument.
Evidence: scripts/generate_network_policy.py allows the creation of executable bash scripts from the args.content parameter without validation or sanitization.
[PROMPT_INJECTION]: The skill exhibits a surface for indirect prompt injection because it processes external data (network policy configurations and content) and provides capabilities to write executable files and run kubectl commands. The absence of strict boundary markers or sanitization for input used in script generation increases the risk of the agent being manipulated into executing unauthorized actions.
Ingestion points: The --content and --config arguments in scripts/generate_network_policy.py accept external data.
Boundary markers: No delimiters or instructions to ignore embedded commands are present in the script generation logic.
Capability inventory: The skill has Write and Edit permissions, along with restricted Bash(kubectl:*) access.
Sanitization: No sanitization or validation is performed on the content before it is written to an executable script.
[SAFE]: The skill references documentation and tools from well-known technology services and official project repositories.
Evidence: Links to kubernetes.io, tigera.io (Calico), and cilium.io are documented as resources.

managing-network-policies