The Agent Skills Directory

[PROMPT_INJECTION]: The skill contains instructions for routing tasks between CLI commands and MCP tools depending on availability. While static analysis flagged potential concealment, the instructions are benign and focused on operational efficiency, emphasizing that the agent should confirm all destructive actions with the user.
[CREDENTIALS_UNSAFE]: The documentation provides clear security boundaries for managing credentials. It explicitly forbids the agent from hardcoding API keys or secrets in logs or commits, and instead recommends using environment variables or official secret management tools. Placeholders used in examples are clearly marked and safe.
[INDIRECT_PROMPT_INJECTION]: The skill exposes capabilities to execute database queries (SQL/NoSQL) and manage file storage, which creates an attack surface for indirect injection. However, it mitigates this by mandating that all generated commands be shown to the user for review and encouraging the use of read-only flags during exploratory queries.
[EXTERNAL_DOWNLOADS]: The skill references standard CLI update and template retrieval mechanisms. These operations fetch resources from established vendor-controlled endpoints (e.g., Tencent Cloud domains) and official package registries, which is consistent with the tool's primary purpose and primary functionality.

cloudbase-cli