literature-review
Pass
Audited by Gen Agent Trust Hub on Jul 24, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructs the agent to execute shell commands using natural language input provided by the user (e.g.,
python scripts/generate_schematic.py "your diagram description"). This creates a risk of command injection if the description contains shell metacharacters and is not correctly escaped by the agent during command construction. - [PROMPT_INJECTION]: The skill ingests untrusted data from academic databases, creating a surface for indirect prompt injection. \n
- Ingestion points: The skill fetches paper titles and abstracts from external APIs including PubMed, arXiv, and Semantic Scholar. \n
- Boundary markers: There are no explicit delimiters or instructions defined to isolate retrieved data from the agent's internal reasoning. \n
- Capability inventory: The skill has access to shell execution via bash and file system operations via write and edit tools. \n
- Sanitization: No sanitization or validation logic is specified for the external metadata before it is processed by the agent.
- [EXTERNAL_DOWNLOADS]: The skill documentation references the installation of standard academic software like Pandoc and LaTeX, and the 'requests' Python library. These are well-known, legitimate tools from official sources and are necessary for the skill's document generation and API interaction capabilities.
Audit Metadata