prompt-engineering-patterns
Pass
Audited by Gen Agent Trust Hub on May 7, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
- [SAFE]: The skill is an educational resource and toolkit for prompt engineering. Analysis of the provided instructions, reference documents, and the optimization script found no malicious code, obfuscation, or unauthorized data access patterns.
- [EXTERNAL_DOWNLOADS]: The skill identifies several standard third-party libraries for AI development, such as
anthropic,openai,langchain,pydantic,numpy, andscikit-learn. These are industry-standard packages from trusted organizations and well-known services. - [PROMPT_INJECTION]: The skill defines a surface for indirect prompt injection as it facilitates processing untrusted data (e.g., user-provided queries, article text, and external context) within LLM prompts. Ingestion points are located in
SKILL.mdand several reference files. The skill mitigates this risk by incorporating boundary markers (e.g., context-restricting instructions inassets/prompt-template-library.md) and sanitization techniques like structured output enforcement via Pydantic inSKILL.md. The skill's capabilities are limited to LLM API calls and local result logging inscripts/optimize-prompt.py.
Audit Metadata