regex-vs-llm-structured-text
Pass
Audited by Gen Agent Trust Hub on Jul 4, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill processes untrusted text through an LLM validation pipeline, creating a surface for indirect prompt injection. \n
- Ingestion points: The
process_documentfunction inSKILL.mdaccepts arbitrary strings (content) which are eventually passed to thevalidate_with_llmfunction. \n - Boundary markers: The prompt in
validate_with_llmuses basic labeling (e.g.,Text: {original_text}) to separate untrusted data from instructions, but lacks more robust delimiters or "ignore embedded instructions" warnings. \n - Capability inventory: The code is limited to data extraction and returning a
ParsedItemobject; it does not execute code, write files, or perform unauthorized network operations based on the parsed data. \n - Sanitization: No input sanitization or output validation is shown in the provided examples before or after LLM processing.
Audit Metadata