hypogenic
Pass
Audited by Gen Agent Trust Hub on Sep 25, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted tabular data and feature strings through large language model prompts, creating a surface for potential indirect prompt injection attacks.\n
- Ingestion points: Training, validation, and test datasets specified in references/config_template.yaml and processed via the BaseTask class.\n
- Boundary markers: Prompt templates in references/config_template.yaml (such as inference and batched_generation) interpolate data directly into {data_samples} and {sample_text} placeholders without using protective delimiters or instructions to ignore embedded commands.\n
- Capability inventory: The skill utilizes command-line interfaces (hypogenic_generation, hypogenic_inference) and a Python API to execute LLM calls and process outputs.\n
- Sanitization: There is no evidence of input validation, escaping, or sanitization for the dataset content before it is included in prompts.\n- [EXTERNAL_DOWNLOADS]: Fetches example datasets and supplemental tools from the ChicagoHAI research laboratory's official GitHub repositories (ChicagoHAI/HypoGeniC-datasets, ChicagoHAI/Hypothesis-agent-datasets, and ChicagoHAI/hypothesis-generation).
Audit Metadata