The Agent Skills Directory

[Data Exposure & Exfiltration] (LOW): The skill performs network requests to reactome.org, which is the legitimate external host for the Reactome database but is not on the predefined whitelist.
[Indirect Prompt Injection] (LOW): The skill ingests external data from files and API responses, creating a surface for potential injection, though no malicious behavior was observed.
Ingestion points: command_analyze in reactome_query.py reads local text files; API responses from https://reactome.org.
Boundary markers: Absent.
Capability inventory: File system write (writing JSON results) and network access via the requests library.
Sanitization: Absent; the agent receives and processes raw data from the API and local files.

reactome-database