cloudbase

Pass

Audited by Gen Agent Trust Hub on Nov 19, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill refers to official Tencent CloudBase resources, directing users to install SDKs and CLI tools via npm (@cloudbase/js-sdk, @cloudbase/node-sdk) and pip (cloudbase-agent-*). It also provides instructions to download official templates and plugin artifacts from github.com/TencentCloudBase and static.cloudbase.net, which are verified vendor-owned sources.
  • [COMMAND_EXECUTION]: The documentation provides comprehensive usage patterns for the tcb CLI, including automated deployment and resource management. The use of flags such as --yes and --force is documented for CI/CD and non-interactive environments, which is standard practice for this vendor's toolchain and does not indicate malicious concealment.
  • [SAFE]: The skill includes mandatory security protocols for developers, such as the 'Sensitive Runtime Data Protection' and 'Change Safety Protocol'. These guides explicitly instruct agents and developers to avoid leaking credentials (like x-cloudbase-context or environment variables) in HTTP responses and to obtain confirmation before destructive changes. It also includes a client-side linting script (cloudbase-lint.mjs) provided as a utility to detect common misconfigurations.
  • [PROMPT_INJECTION]: Static detections for prompt injection were found to be benign. The reference to 'ignore previous instructions' in references/ops-inspector/SKILL.md is a defensive instruction to the agent to treat log data as untrusted text rather than executable commands. Detections related to concealment are associated with standard non-interactive CLI arguments used for legitimate deployment workflows.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 19, 2026, 07:23 AM