The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill's primary function is to review and modify external, potentially untrusted container artifacts including Dockerfiles, Kubernetes manifests, and Helm charts.
Ingestion points: The 'review-a-kubernetes-manifest' and 'write-or-slim-a-container-image' workflows in SKILL.md process external file content directly into the agent's context.
Boundary markers: The instructions do not specify the use of delimiters or 'ignore embedded instructions' warnings when processing these external files.
Capability inventory: The skill leverages powerful CLI tools including docker, kubectl, and helm for resource inspection and validation.
Sanitization: There is no evidence of automated content sanitization or filtering for the artifacts being reviewed.
[COMMAND_EXECUTION]: The skill contains a wide array of command-line instructions for docker, kubectl, helm, and kubeconform to be used for diagnostics, troubleshooting, and validation of containerized environments.
[CREDENTIALS_UNSAFE]: Multiple files within the evals/files/ directory contain hardcoded placeholder credentials intended for use as test cases in security auditing evaluations.
Evidence: evals/files/deployment.yaml contains a hardcoded DB_PASSWORD ('s3cr3t-prod-password') and a dummy Stripe API key ('sk_live_51H8xQwEXAMPLEEXAMPLE').
Context: These credentials are documented as security vulnerabilities within the skill's own rules (Rule 21 in SKILL.md) and are provided solely to verify the agent's detection capabilities during testing.

containers