The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill includes a 'Diagnosis Workflow' designed to analyze error patterns by retrieving system and user logs via the get-log command. As these logs are generated by external task execution environments and are processed by the agent to generate fix suggestions, they represent an ingestion surface for untrusted data. There are no explicit boundary markers or sanitization steps mentioned to prevent the agent from following malicious instructions potentially embedded in those logs.
[CREDENTIALS_UNSAFE]: The documentation follows safe practices by instructing users to configure access keys (HUAWEI_CLOUD_AK and HUAWEI_CLOUD_SK) through environment variables or a configuration file (~/.cloudrobo/config.yaml). It explicitly warns that credentials must never be hardcoded in scripts or documents.
[EXTERNAL_DOWNLOADS]: The skill documentation references and requires the hw-cloudrobo-client Python package, which is the official vendor library for interacting with the CloudRobo service. These references are documented neutrally as standard requirements for the skill's operation.

huawei-cloud-cloudrobo-dataset