regex-vs-llm-structured-text
Pass
Audited by Gen Agent Trust Hub on Mar 12, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill exhibits a surface for indirect prompt injection in the
validate_with_llmfunction. - Ingestion points: The
contentparameter in theprocess_documentfunction (SKILL.md) is the primary entry point for untrusted data. - Boundary markers: Absent. The LLM prompt in
validate_with_llminterpolates theoriginal_textdirectly using only basic text labels, which may allow embedded instructions to influence the model's behavior. - Capability inventory: The skill does not possess high-risk capabilities such as file system access, arbitrary command execution, or network requests beyond the intended LLM call.
- Sanitization: Absent. There is no evidence of input validation, escaping, or structural filtering of the processed text before it is interpolated into the LLM prompt.
Audit Metadata