reactome-database
Warn
Audited by Gen Agent Trust Hub on Jun 27, 2026
Risk Level: MEDIUMDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
- [DATA_EXFILTRATION]: The
scripts/reactome_query.pyscript accepts a file path as a command-line argument and transmits the file's content to the Reactome Analysis Service (https://reactome.org/AnalysisService/identifiers/) via HTTP POST. This design enables the potential exfiltration of sensitive local data, such as.envfiles or credentials, if the agent is directed to target paths outside of the intended biological datasets.\n- [PROMPT_INJECTION]: The skill processes untrusted biological identifiers and expression data from external files. These inputs could contain malicious instructions designed to influence the agent's behavior when it parses and presents the analysis results.\n - Ingestion points: Gene list files processed by
scripts/reactome_query.pyand expression data inputs described inSKILL.md.\n - Boundary markers: The skill does not implement delimiters or 'ignore' instructions to isolate data from potential embedded commands.\n
- Capability inventory: The skill performs network operations to the Reactome API and writes output to the local filesystem.\n
- Sanitization: Input data is only processed with basic whitespace stripping (
strip()), lacking structural validation or content filtering to prevent injection.
Audit Metadata