regex-vs-llm-structured-text
Pass
Audited by Gen Agent Trust Hub on Apr 7, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill implements a hybrid parsing pipeline where untrusted input text (
original_text) is passed directly into an LLM prompt in thevalidate_with_llmfunction. This creates a surface for indirect prompt injection, where an attacker could embed malicious instructions in the text being parsed to manipulate the LLM's validation logic.\n - Ingestion points: The
contentparameter inprocess_documentand theoriginal_textparameter invalidate_with_llm(SKILL.md).\n - Boundary markers: Absent. Input text is concatenated with instructions using f-strings without delimiters (e.g., XML tags or backticks) or instructions to ignore embedded commands.\n
- Capability inventory: The skill performs LLM invocations using the
client.messages.createmethod with the 'claude-haiku-latest' model.\n - Sanitization: No escaping, sanitization, or validation of the input text is performed before it is included in the prompt.
Audit Metadata