faion-llm-integration
Pass
Audited by Gen Agent Trust Hub on Feb 17, 2026
Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- Prompt Injection (LOW): The skill utilizes prompt templates (e.g., in
prompt-basics/README.md) that interpolate user-provided data into instructions, creating a surface for indirect prompt injection attacks.\n - Evidence Chain (Category 8):\n
- Ingestion points: User data is interpolated into prompt templates in
prompt-basics/README.mdandopenai-api-integration/README.md.\n - Boundary markers: The use of delimiters (e.g.,
---) is suggested in theprompt-basics/README.mddocumentation but not strictly enforced across all examples.\n - Capability inventory: The skill is configured with powerful tools including
Bash,Write, andEditas specified inSKILL.md.\n - Sanitization: Robust mitigation patterns and code for content moderation and injection detection are provided as core methodology in
guardrails-basics/README.md.\n- Command Execution (LOW): The skill is permitted to use theBashandTasktools as configured inSKILL.md. While these are high-risk capabilities, they are consistent with the primary purpose of a Machine Learning engineering skill and are accompanied by safety best practices and guardrail documentation.
Audit Metadata