deepstream-inspect-stack
Pass
Audited by Gen Agent Trust Hub on Nov 6, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill provides a controlled environment for diagnosing DeepStream version mismatches and installation errors. It incorporates safety-first instructions, such as rerouting requests for code generation or model importing to specialized sibling skills, and mandates verification against official documentation.\n- [COMMAND_EXECUTION]: The skill defines an autonomous diagnostic mode (Mode B) that uses read-only shell commands (e.g.,
nvidia-smi,dpkg,cat,docker ps) to retrieve system metadata. This mode is governed by strict constraints: commands must be announced before execution, state-changing operations are forbidden, and user opt-in is required.\n- [DATA_EXFILTRATION]: The skill instructs the agent to redact sensitive information—including private registry URLs and internal build identifiers—from command outputs. This proactive measure prevents the exposure of proprietary configuration data in logs or agent responses.\n- [EXTERNAL_DOWNLOADS]: All external resources referenced in the skill are official NVIDIA documentation sites or public catalogs, ensuring that all data sources are trusted and authenticated.
Audit Metadata