Zotero
Pass
Audited by Gen Agent Trust Hub on Nov 1, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- Application Management via Subprocess: The skill utilizes the Python
subprocessmodule to manage the Zotero desktop application, including operations like restarting the process to apply preference changes. These commands are targeted at specific system utilities (such asosascript,taskkill, orpkill) to support the skill's primary function of ensuring the local API is active. - File System Interaction: The helper script accesses and modifies local files, specifically Zotero's preference files and user-provided research documents (Markdown or LaTeX). These actions are necessary to enable the Zotero local API and automate citation insertion, which are core features of the skill.
- Indirect Prompt Injection Surface: The skill processes information retrieved from the local Zotero library, such as paper titles, abstracts, and full-text content. This creates a potential surface where instructions embedded in the library data could be processed by the agent.
- Ingestion points: Library data is fetched from the Zotero local API through the
scripts/zotero.pyscript. - Boundary markers: The skill does not currently define explicit delimiters or instructions to separate external library data from the agent's internal instructions.
- Capability inventory: The skill possesses capabilities to write to the local file system and manage the Zotero application process.
- Sanitization: The retrieved content is processed directly without specific sanitization or filtering for embedded instructions.
Audit Metadata