food-deep-research
Warn
Audited by Gen Agent Trust Hub on Jul 26, 2026
Risk Level: MEDIUMCOMMAND_EXECUTIONDATA_EXFILTRATIONPROMPT_INJECTIONCREDENTIALS_UNSAFE
Full Analysis
- [COMMAND_EXECUTION]: The skill instructs the agent to run several local Python scripts (e.g., scripts/food_fetch_setup.py, scripts/verify_citations.py). Because the source code for these scripts is not included in the provided files, their behavior cannot be verified, presenting a risk of unauthorized command execution.
- [DATA_EXFILTRATION]: The skill requests access to personal citation manager data folders, such as EndNote and Zotero. While used for legitimate research purposes, this provides a broad surface for sensitive data exposure and potential exfiltration.
- [PROMPT_INJECTION]: The skill processes external data from web searches and PDF files, creating an Indirect Prompt Injection surface. It lacks explicit boundary markers or instructions to prevent the agent from acting on malicious commands hidden within external research documents.
- [CREDENTIALS_UNSAFE]: The skill prompts the user for institutional access to reach paywalled journals. There is no specified secure mechanism for handling these authentication details, which may lead to the exposure of library credentials.
Audit Metadata