research-assistant
Pass
Audited by Gen Agent Trust Hub on Apr 13, 2026
Risk Level: SAFEPROMPT_INJECTIONNO_CODECOMMAND_EXECUTION
Full Analysis
- [SAFE]: The skill implements a well-defined academic workflow that manages research data and drafts within a localized
_workspace/directory. It uses standard academic resources (RISS, DBpia) for literature searching. - [PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface because it processes untrusted content from external academic databases and user-provided manuscripts.
- Ingestion points: Academic database search results (RISS, DBpia, Google Scholar, PubMed) and user-uploaded documents stored in the
_workspace/folder. - Boundary markers: Absent. The instructions do not mandate the use of delimiters or 'ignore' instructions when processing external literature.
- Capability inventory: The skill can perform file-write operations within the workspace and utilize the sequential thinking tool for complex reasoning.
- Sanitization: No explicit sanitization or instruction-filtering is applied to the retrieved academic content.
- [COMMAND_EXECUTION]: The skill includes a 'Statistical Analyst' role responsible for generating R or Python code for data analysis. While the skill describes the generation of this code rather than its direct execution on the host, users should review any generated scripts before manual execution.
Audit Metadata