ultimate-search
Pass
Audited by Gen Agent Trust Hub on Mar 7, 2026
Risk Level: SAFE
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill uses Docker Compose to deploy several helper services including FlareSolverr, grok2api, and TavilyProxyManager. These are known community tools used for API aggregation and web automation, and their use is well-documented in the skill's README.
- [DATA_EXFILTRATION]: The skill handles sensitive API keys and session tokens through local .env and text files. These credentials are transmitted only to official API endpoints or local service proxies. No unauthorized data transmission to third-party domains was detected.
- [COMMAND_EXECUTION]: Local bash scripts serve as the primary interface for search operations. These scripts utilize
jqfor input validation and JSON construction, ensuring that untrusted data from search results does not interfere with the execution environment. - [PROMPT_INJECTION]: The skill mitigates risks associated with processing external web data (Indirect Prompt Injection) by including mandatory verification rules in the SKILL.md. These instructions explicitly direct the agent to treat search results as untrusted suggestions and require cross-referencing facts against multiple sources.
- [SAFE]: The project demonstrates a mature approach to security, including the use of SSH tunneling for remote management and binding internal services strictly to the loopback interface (127.0.0.1).
Audit Metadata