building-neural-networks
Pass
Audited by Gen Agent Trust Hub on Mar 12, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill exhibits an attack surface for indirect prompt injection where malicious instructions could be embedded in data processed by the agent.
- Ingestion points: The scripts
scripts/build_network.py,scripts/train_network.py,scripts/evaluate_network.py, andscripts/visualize_network.pyaccept and process arbitrary input files and directories. - Boundary markers: There are no delimiters or instructions provided to the agent to ignore potentially malicious instructions embedded within the processed neural network configurations or datasets.
- Capability inventory: The skill is granted broad permissions in
SKILL.md, specificallyBash(cmd:*), allowing for potentially dangerous command execution if an injection occurs. - Sanitization: Data read from external files (especially JSON) is processed without validation or sanitization before entering the agent's context.
Audit Metadata