The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill represents a surface for indirect prompt injection as it ingests and processes untrusted data from external sources.
Ingestion points: Untrusted data enters the agent context through command outputs such as cloudrobo train get-logs, cloudrobo train get-events, and cloudrobo train show-task, as well as user-supplied configurations in create-task.
Boundary markers: The instructions do not explicitly mandate the use of boundary markers or delimiters when presenting this external content to the LLM.
Capability inventory: The skill possesses capabilities to execute shell commands (cloudrobo train), modify cloud resources (create-task, update-task), and perform lifecycle operations (stop-task, delete-tasks).
Sanitization: There is no mention of explicit sanitization or filtering of the ingested logs or event messages before they are processed by the agent.
[SAFE]: The skill uses official Huawei Cloud resources and follows security best practices for credential management.
Network operations: The skill communicates with official Huawei Cloud domains (e.g., cloudrobo.cn-southwest-2.myhuaweicloud.com) for API interactions.
Credential management: The skill correctly instructs users to provide credentials via environment variables (HUAWEI_CLOUD_AK, HUAWEI_CLOUD_SK) or configuration files, and explicitly warns against hardcoding them in scripts.
Dependencies: The skill references the official vendor-owned Python package hw-cloudrobo-client for its operations.

huawei-cloud-cloudrobo-train