ai-parsing-data
Pass
Audited by Gen Agent Trust Hub on May 13, 2026
Risk Level: SAFE
Full Analysis
- [PROMPT_INJECTION]: The skill is designed to process untrusted external data (such as emails, call transcripts, and Langfuse traces), which is an inherent surface for indirect prompt injection. The skill correctly mitigates this risk by demonstrating structural parsing through DSPy Signatures and Pydantic models, which serve as boundary markers for the AI. It also includes validation steps using assertions to verify the format of extracted data.
- Ingestion points: Processes text from various sources including local text files (
document.txt), CSV files (emails.csv), and external API traces (Langfuse). - Boundary markers: Uses structural delimiters via DSPy signatures and Pydantic models to separate input data from instructions.
- Capability inventory: Limited to file system operations for reading/writing data and network access via the Langfuse SDK for observability.
- Sanitization: Implements field-level validation and structural constraints via Pydantic types and DSPy assertions (
dspy.Suggest). - [EXTERNAL_DOWNLOADS]: The skill references and uses standard, well-known Python libraries including
dspy,pydantic,pandas, andlangfuse. These are established tools in the AI development ecosystem and are used here for their intended purposes.
Audit Metadata