prompt-engineering-patterns

Pass

Audited by Gen Agent Trust Hub on Mar 28, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [SAFE]: The skill is primarily educational, providing well-documented patterns for few-shot learning, chain-of-thought prompting, and prompt optimization.\n- [SAFE]: Code snippets in the references and the optimization script use standard Python libraries such as numpy, scikit-learn, and sentence-transformers for utility purposes like semantic similarity and metric calculation.\n- [SAFE]: No evidence of hardcoded credentials, malicious network operations, or obfuscation was found across the analyzed files.\n- [PROMPT_INJECTION]: The skill implements prompt template systems and optimization scripts that interpolate variables into prompts, which constitutes an indirect prompt injection surface.\n
  • Ingestion points: Ingestion of user-provided variables occurs in PromptOptimizer.evaluate_prompt (scripts/optimize-prompt.py) and PromptTemplate.render (references/prompt-templates.md).\n
  • Boundary markers: No explicit boundary markers or instructions to ignore embedded commands are present in the rendering logic.\n
  • Capability inventory: Interpolated prompts are executed via an LLM client (e.g., client.complete or openai.ChatCompletion.create) across multiple reference scripts.\n
  • Sanitization: There is no evidence of input sanitization, escaping, or validation of interpolated variables before they are sent to the model.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 28, 2026, 08:18 PM
Security Audit — agent-trust-hub — prompt-engineering-patterns