hypogenic
Pass
Audited by Gen Agent Trust Hub on Apr 16, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: Fetches example datasets from the 'ChicagoHAI' GitHub organization and installs the 'hypogenic' package from the official Python Package Index (PyPI).
- [COMMAND_EXECUTION]: Instructs the user to execute local shell scripts (
setup_grobid.sh,run_grobid.sh) provided within the skill's directory structure to manage environment setup and background services. - [PROMPT_INJECTION]: Identified an indirect prompt injection surface (Category 8) because the skill processes external datasets and literature PDFs to generate prompts.
- Ingestion points: Training, validation, and test datasets in JSON format, and research paper PDFs.
- Boundary markers: Uses YAML configuration templates with
${variable}placeholders to interpolate data into system and user prompts. - Capability inventory: Includes CLI tools for generation and inference (
hypogenic_generation,hypogenic_inference) and python scripts for PDF processing. - Sanitization: Relies on structured template injection and custom label extraction functions to parse LLM outputs.
Audit Metadata