The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill provides instructions and commands to install the aliyun CLI using a script from https://aliyuncli.alicdn.com/setup.sh. Since this URL belongs to the official Alibaba Cloud CDN and is used to install the vendor's own tooling, it is considered a legitimate and safe administrative operation.
[COMMAND_EXECUTION]: The skill executes aliyun CLI and kubectl commands via subprocess.run. All such calls in the provided scripts (e.g., common.py, test_m2m_token.py) use list-formatted arguments rather than shell strings, which is a security best practice that prevents command injection.
[DATA_EXPOSURE_AND_EXFILTRATION]: The skill includes explicit safeguards against data exposure. The scripts/test_m2m_token.py script automatically redacts sensitive fields like access_token, refresh_token, and id_token from its output. Additionally, SKILL.md contains strict rules prohibiting the agent from reading or printing Access Keys, directing the user to manage credentials through the official CLI configuration instead.
[INDIRECT_PROMPT_INJECTION]: The skill processes external data from Kubernetes clusters, such as ServiceAccount tokens and OIDC discovery JSON. The ingestion is handled by dedicated Python parsing scripts (parse_sa_token.py, probe_k8s.py), which minimizes the risk of embedded malicious instructions influencing the agent's behavior compared to direct prompt interpolation.

alibabacloud-idaas-kubernetes-m2m-federated-credential