hypogenic
Pass
Audited by Gen Agent Trust Hub on Jun 26, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: Fetches and installs the 'hypogenic' package from the official Python Package Index (PyPI) and clones research datasets from GitHub repositories managed by ChicagoHAI.
- [COMMAND_EXECUTION]: Executes shell scripts ('setup_grobid.sh' and 'run_grobid.sh') to initialize and run the GROBID service, which is used for processing scientific literature in PDF format.
- [PROMPT_INJECTION]: The skill implements a workflow that ingests external tabular data and interpolates it into LLM prompt templates, which constitutes an indirect prompt injection surface.
- Ingestion points: Data features from JSON files ('text_features_1' through 'text_features_n') specified in the task configuration.
- Boundary markers: The skill does not employ explicit boundary markers or 'ignore' instructions around interpolated data in its default templates.
- Capability inventory: The framework performs network operations (API calls to LLMs) and file system access (reading datasets and writing hypothesis banks).
- Sanitization: No evidence of input sanitization or validation was found for the feature values ingested from datasets.
Audit Metadata