regex-vs-llm-structured-text
Pass
Audited by Gen Agent Trust Hub on Apr 14, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill defines a hybrid pipeline that ingests untrusted text and interpolates it into an LLM prompt, creating a surface for indirect prompt injection. \n
- Ingestion points: The
process_documentfunction inSKILL.mdaccepts raw string content from potentially untrusted sources (e.g., uploaded documents or forms).\n - Boundary markers: The prompt template in the
validate_with_llmfunction lacks explicit delimiters (such as XML tags or triple backticks) to isolate the untrustedoriginal_textfrom the agent's instructions.\n - Capability inventory: No high-risk capabilities, such as shell command execution (
subprocess), file system writes, or arbitrary network exfiltration, were detected in the skill's code snippets.\n - Sanitization: The implementation does not include logic to sanitize or escape the input text before it is processed by the LLM validator.
Audit Metadata