regex-vs-llm-structured-text
Pass
Audited by Gen Agent Trust Hub on Mar 4, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection within its validation logic.
- Ingestion points: The
validate_with_llmfunction inSKILL.mdacceptsoriginal_text, which is untrusted data from a document being processed. - Boundary markers: The prompt template uses simple text headers ("Text: {original_text}") rather than secure delimiters or explicit instructions to the LLM to ignore instructions found within the data.
- Capability inventory: While the provided code only returns structured data, an agent using this skill to parse instructions or data for subsequent tools could be manipulated if the LLM's output is poisoned by the input text.
- Sanitization: There is no evidence of sanitization, escaping, or structural validation of the input text before it is interpolated into the LLM prompt.
Audit Metadata