modelopt-run-agentx
Pass
Audited by Gen Agent Trust Hub on Nov 13, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill downloads and installs the
agentx-harnesspackage from PyPI, a well-known package registry. - [EXTERNAL_DOWNLOADS]: References documentation and code from the
SemiAnalysisAI/agentx-harnessrepository on GitHub. - [COMMAND_EXECUTION]: The skill executes shell commands to create a virtual environment, install dependencies, and run the
aiperfbenchmarking tool. - [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from external Hugging Face datasets (
AGENTX_DATASET) and interacts with potentially untrusted model endpoints. It mitigates these risks by instructing the user to pin dataset revisions using commit SHAs and requiring explicit user consent before enabling remote code for tokenizers.
Audit Metadata