The Agent Skills Directory

[COMMAND_EXECUTION]: The skill requires executing several shell commands for installation and operation, including package installation via uv pip, repository cloning via git, and running internal setup scripts like setup_grobid.sh and run_grobid.sh.\n- [EXTERNAL_DOWNLOADS]: The skill downloads datasets and code from external sources including GitHub repositories (ChicagoHAI) and the PyPI registry for the hypogenic package.\n- [PROMPT_INJECTION]: The skill exhibits an attack surface for indirect prompt injection during its hypothesis generation and inference workflows.\n
Ingestion points: The skill processes untrusted data from user-provided JSON datasets and research literature PDFs.\n
Boundary markers: Prompt templates in the configuration use direct interpolation (e.g., {sample_text}) without implementing strong boundary delimiters or instructions to ignore embedded commands.\n
Capability inventory: The ingested data is used to drive LLM-based hypothesis generation, refinement, and validation via the hypogenic API and CLI.\n
Sanitization: No specific mechanisms for input sanitization or validation are provided to filter out instructions that may be embedded within datasets or literature.

hypogenic