chi-reproducibility
Pass
Audited by Gen Agent Trust Hub on Oct 7, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill includes shell commands to automate research verification. It validates the presence of mandatory project directories, searches for weak availability statements within the paper text, and attempts to run the analysis pipeline by creating a virtual environment and installing dependencies from a local requirements file.
- [INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied research documents and datasets, creating a surface for indirect prompt injection.
- Ingestion points: Content within the
paper/,data/, andanalysis/directories. - Boundary markers: None identified; instructions are interpreted directly from the environment.
- Capability inventory: Includes file system listing (
ls), text searching (grep), package installation (pip install), and Python script execution (python). - Sanitization: No explicit sanitization or filtering of the processed research files is performed.
Audit Metadata