The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill possesses a theoretical attack surface because it ingests and processes pod configuration data from an external Kubernetes API.
Ingestion points: Pod specifications are fetched from the cluster using the v1.list_pod_for_all_namespaces() method in scripts/agent.py.
Boundary markers: The script does not implement specific delimiters or warnings to ignore embedded instructions within the processed pod metadata.
Capability inventory: The agent script is capable of performing network operations (Kubernetes API requests) and writing report data to a local file (container_escape_report.json).
Sanitization: The script does not perform sanitization on external strings such as pod names or namespace identifiers before they are included in the generated audit report.
[EXTERNAL_DOWNLOADS]: The skill utilizes the standard kubernetes Python client to interact with cluster environments.
Evidence: The requirement for the kubernetes library is documented in SKILL.md and scripts/agent.py. This is an official, well-known dependency expected for the skill's stated purpose of auditing Kubernetes security posture.

performing-container-escape-detection