academic-deep-research
Pass
Audited by Gen Agent Trust Hub on Feb 23, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: No malicious behavior, obfuscation, or safety bypass attempts were detected within the skill instructions or reference files. The skill is entirely self-contained, relying on the native capabilities of the agent's environment without requiring external API keys or third-party dependencies.\n- [PROMPT_INJECTION]: While the skill retrieves untrusted data from the internet via web search and fetch tools, it is not considered a high-risk vector due to the skill's design. The protocol mandates 'explicit analysis between every tool call' and multiple research cycles, which act as a logical buffer against indirect prompt injection. Furthermore, the agent is restricted to informational tools and lacks the system-level permissions required to execute malicious commands found in external content.\n- [EXTERNAL_DOWNLOADS]: The skill utilizes standard web search and fetch operations (web_search, web_fetch) as part of its core research function. These activities are limited to the user's research topic and occur through the platform's native tools, posing no risk of unauthorized software installation or system compromise.
Audit Metadata