llm-serving-capacity-planner
Pass
Audited by Gen Agent Trust Hub on Sep 1, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes external, untrusted log files and model configuration files as its primary input. This creates an attack surface where maliciously crafted content within these files could attempt to influence the AI agent's behavior when it reviews the resulting analysis report.\n
- Ingestion points: The
scripts/capacity_analyzer.pytool ingests data via the--log-file,--nvidia-smi-file, and--config-jsoncommand-line arguments.\n - Boundary markers: The output generated by the skill consists of plain text tables and tuning suggestions without explicit delimiters or warnings to the AI agent to ignore instructions that might be embedded in the log data.\n
- Capability inventory: The analysis script is limited to parsing and reporting; it does not contain tools for network requests, file writing, or command execution.\n
- Sanitization: While the script uses regular expressions to extract specific numerical values, it does not include general sanitization for all text extracted from the logs (such as model paths).
Audit Metadata