profile-isaac-sim
Warn
Audited by Gen Agent Trust Hub on Jul 30, 2026
Risk Level: MEDIUMCOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill's workflow for profiling extension unit tests directs the agent to capture a command line from the 'running process:' string in console logs and execute it directly in the shell. This dynamic command discovery and execution is a sensitive pattern that relies on the integrity of the log source.
- [EXTERNAL_DOWNLOADS]: Fetches assets and configurations from an official Amazon S3 bucket used for NVIDIA Omniverse content staging to bypass interactive authentication during benchmarks.
- [PROMPT_INJECTION]: The skill possesses an attack surface for indirect prompt injection by ingesting and parsing untrusted data from external Tracy CSV exports and benchmark log files.
- Ingestion points: Metric extraction from log files in
SKILL.mdand CSV data processing inscripts/compare_tracy_csvs.py. - Boundary markers: No explicit markers or delimiters are defined to separate the data from instructions during processing.
- Capability inventory: The agent can execute several subprocess commands including
python.sh,tracy-capture, and local scripts. - Sanitization: There is no evidence of input validation or content sanitization for the performance data being ingested.
Audit Metadata