The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/tke.py and example code in SKILL.md execute the kubectl command-line tool using subprocess.run. The implementation correctly uses argument lists rather than raw shell strings, which mitigates standard shell injection vulnerabilities while allowing the agent to perform cluster operations.
[EXTERNAL_DOWNLOADS]: The skill identifies dependencies on the official tencentcloud-sdk-python library and the kubectl binary. These are standard, well-known tools required for managing Tencent Cloud infrastructure.
[INDIRECT_PROMPT_INJECTION]: The skill provides functions to read Kubernetes resource metadata, events, and pod logs. These external data sources are untrusted and could contain malicious instructions intended to mislead the agent during troubleshooting or management tasks.
Ingestion points: Untrusted data enters the agent context via kubectl logs, kubectl describe, and kubectl get events commands in scripts/tke.py.
Boundary markers: No explicit delimiters or instructions to ignore embedded commands are present when processing tool output.
Capability inventory: The skill has the ability to write sensitive kubeconfig files and execute cluster-modifying commands such as scale and rollout restart.
Sanitization: There is no evidence of filtering or sanitizing the content of logs or resource descriptions before they are processed by the agent.

tencentcloud-tke