The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes a local Python script scripts/generate.py to produce Kubernetes manifests. While this is part of the skill's functionality, it processes user-supplied arguments (storage and image) which leads to further vulnerabilities.
[PROMPT_INJECTION]: The scripts/generate.py script is vulnerable to indirect injection. User-provided values for storage_size and image are interpolated directly into a multi-line f-string without sanitization or validation. An attacker could provide crafted input (e.g., using YAML document separators --- or additional fields) to inject arbitrary Kubernetes resources or modify the security context of the deployment.
Ingestion points: User-provided storage and image parameters in SKILL.md.
Boundary markers: Absent.
Capability inventory: The skill produces YAML manifests intended for deployment to a cluster.
Sanitization: Absent; the script uses direct string interpolation.
[COMMAND_EXECUTION]: The generated manifest explicitly requires hostNetwork: true. This is a high-risk security configuration in Kubernetes that allows the database container to share the host's network namespace, bypassing container isolation and potentially exposing host-level services to the database process.
[CREDENTIALS_UNSAFE]: The generation script includes the randomly generated password in a plaintext YAML comment (# Note: Raw password for reference). This practice increases the risk of credential exposure if the generated manifests are stored in logs, version control, or shared across insecure channels.

postgre-skill