model-compute-simulation
Pass
Audited by Gen Agent Trust Hub on Oct 25, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill implements a performance analysis toolset for transformer models. All operations are performed using local Python scripts and static configuration files located within the skill's directory.
- [COMMAND_EXECUTION]: The documentation provides standard instructions for the agent to invoke local scripts (
model_compute_simulator.pyandextract_compute_flow_from_trace.py) with specific CLI arguments. These commands are used for deterministic calculations and do not involve shell injection vulnerabilities. - [INDIRECT_PROMPT_INJECTION]: The skill contains logic to process external data sources such as HuggingFace
config.jsonfiles and torch profiler traces. The scripts perform strict structured parsing (JSON/GZIP) for architectural parameters (e.g., hidden size, number of layers) and do not interpret or execute instructions embedded within the processed data.
Audit Metadata