nlp-processing
Pass
Audited by Gen Agent Trust Hub on Sep 29, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill utilizes the
transformerslibrary andspacyto load pre-trained models (e.g., BERT, GPT-2) from official and well-known registries. These downloads are standard for NLP tasks and originate from established services in the AI ecosystem.\n- [INDIRECT_PROMPT_INJECTION]: The skill contains logic for processing external, potentially untrusted text data, creating a surface for indirect prompt injection.\n - Ingestion points: Data enters through the
TextPreprocessorclass inscripts/text_preprocessor.pyand the preprocessing functions inSKILL.md.\n - Boundary markers: The code snippets do not implement specific boundary markers for LLM prompts.\n
- Capability inventory: The skill includes capabilities to write training outputs to the local filesystem and generate text based on input.\n
- Sanitization: The skill features significant mitigation through its
TextPreprocessorscript, which removes HTML tags, URLs, email addresses, and extra whitespace, effectively sanitizing input before it is processed by downstream models.
Audit Metadata