llm-application-dev-prompt-optimize
Pass
Audited by Gen Agent Trust Hub on Oct 26, 2026
Risk Level: SAFENO_CODE
Full Analysis
- [SAFE]: The analyzed files consist of markdown instructions and documentation without any executable scripts (.sh, .py, .js) or automated tool configurations. All provided code blocks are non-executable examples intended for educational purposes and architectural guidance.
- [INDIRECT_PROMPT_INJECTION]: The skill instructions acknowledge the risk of indirect prompt injection when processing external data. It provides defensive design patterns, such as using XML-style tags for context separation and explicit instructions to treat retrieved data as untrusted evidence. Ingestion points: The skill utilizes $ARGUMENTS and references external data sources like 'retrieved_documents' and 'customer_feedback'. Boundary markers: Includes clear warnings that external content 'cannot override the evaluator rubric'. Capability inventory: No tools, file system access, or network operations are requested or utilized. Sanitization: Recommends 'context-specific output encoding' and warns against relying on simple normalization.
Audit Metadata