project-guidelines-example
Pass
Audited by Gen Agent Trust Hub on Sep 1, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The example code for Claude AI integration demonstrates a vulnerability where untrusted data could influence agent behavior.
- Ingestion points: The
analyze_with_claudefunction inSKILL.mdaccepts acontentstring from an unspecified external source. - Boundary markers: The example code lacks any delimiters or specific instructions (e.g., system prompts or XML tags) to help the model distinguish between instructions and the data being analyzed.
- Capability inventory: The model is granted tool-calling capabilities (
provide_analysis) which could be misused if thecontentcontains malicious instructions. - Sanitization: There is no evidence of input validation or sanitization before the data is passed to the AI model.
Audit Metadata