The Agent Skills Directory

[CREDENTIALS_UNSAFE]: In scripts/huawei_cloud/hcloud_runner.py, the run and run_with_body functions pass sensitive Huawei Cloud credentials (AK/SK) as command-line arguments (--cli-access-key, --cli-secret-key) to the hcloud process. This occurs if the credentials are not already configured in the local hcloud config file. Command-line arguments are visible to all users on the host system via process monitoring tools like ps, creating a risk of local credential harvesting.
[EXTERNAL_DOWNLOADS]: The SKILL.md and references/cce-api-guide.md files provide instructions to download and execute an installer script for the hcloud CLI from Huawei Cloud's official domain: https://cn-north-4-hdn-koocli.obs.cn-north-4.myhuaweicloud.com/cli/latest/hcloud_install.sh. This is a vendor-provided tool required for the skill's operation.
[INDIRECT_PROMPT_INJECTION]: The skill ingests data from external CCE APIs that could be manipulated by an attacker to influence agent behavior.
Ingestion points: Metadata for clusters, nodes, and addons is retrieved via tools in scripts/huawei_cloud/dispatcher.py and incorporated into the agent's context.
Boundary markers: The skill does not use specific delimiters or instructions to prevent the agent from obeying natural language instructions that might be embedded in the API responses.
Capability inventory: The skill has high-privilege capabilities, including the ability to delete clusters, delete nodes, and uninstall addons across multiple scripts in scripts/huawei_cloud/special_ops.py.
Sanitization: There is no specific validation or sanitization of string data returned from the cloud API to identify or neutralize potential injection attempts.
[COMMAND_EXECUTION]: The skill executes various system commands using subprocess.run to interact with hcloud and kubectl. While these calls use a secure list-based argument format, they provide the agent with broad control over the cloud environment.

huawei-cloud-cce-cluster-management