optimizing-deep-learning-models

Pass

Audited by Gen Agent Trust Hub on May 20, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill exhibits a surface for indirect prompt injection. It is designed to analyze local files such as model architectures and training data, which could contain malicious instructions or "jailbreak" attempts embedded in comments or metadata within those files.
  • Ingestion points: The skill instructions in SKILL.md specify that the agent reads local model architecture and performance data to determine optimization strategies.
  • Boundary markers: There are no protective boundary markers or instructions to the agent to ignore commands within the analyzed data provided in the skill instructions.
  • Capability inventory: The skill is configured with powerful tools including Bash(cmd:*), Write, and Edit as seen in the YAML frontmatter of SKILL.md, allowing for high-impact system operations.
  • Sanitization: The provided analysis script (scripts/analyze_model.py) focuses on metadata analysis (file size, count) and does not provide sanitization for file content that the agent may read.
Audit Metadata
Risk Level
SAFE
Analyzed
May 20, 2026, 03:13 PM
Security Audit — agent-trust-hub — optimizing-deep-learning-models