prompt-engineering-patterns
Pass
Audited by Gen Agent Trust Hub on Apr 12, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill provides code templates that interpolate external data directly into LLM prompts, creating a surface for indirect prompt injection.
- Ingestion points: Code samples in
SKILL.mduse variables like{query},{text}, and{context}to insert external content into prompts. - Boundary markers: The provided patterns lack robust delimiters or specific instructions to the model to ignore potential commands within the interpolated text.
- Capability inventory: The demonstrated capabilities include interacting with external LLM APIs (Anthropic, VoyageAI) and vector databases.
- Sanitization: The code snippets do not implement input validation or sanitization for the data being placed into the prompt templates.
Audit Metadata