rnacentral-skill
Pass
Audited by Gen Agent Trust Hub on Sep 20, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSDATA_EXFILTRATIONINDIRECT_PROMPT_INJECTION
Full Analysis
- Generic Network Access: The utility script
scripts/rest_request.pyis a flexible REST client that accepts abase_urlfrom the agent's input. Because the script does not validate or restrict this URL to a specific domain, it could potentially be used to interact with any network-accessible endpoint, including internal services or external domains not associated with RNAcentral. - Local File Writing Capability: The skill includes logic to save raw response data to the local file system through the
save_rawandraw_output_pathparameters. While intended for caching or logging, the ability to specify arbitrary file paths could allow the creation or modification of files in the local environment. - Indirect Prompt Injection Surface: The skill processes external data from biological databases which are then summarized for the agent. This represents a potential vector where instructions embedded in external data could influence the agent's behavior.
- Ingestion points: Data enters the environment from external REST API responses in
scripts/rest_request.py. - Boundary markers: The script returns data in a structured JSON format, which provides clear separation between the tool's status and the retrieved content.
- Capability inventory: The script performs network requests via the
requestslibrary and can write files to disk using thepathlibmodule. - Sanitization: The script performs string truncation and depth-limiting via a
_compactfunction to manage payload size, but it does not perform semantic safety filtering on the retrieved content.
Audit Metadata