zot
Pass
Audited by Gen Agent Trust Hub on Sep 16, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONDATA_EXFILTRATIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from PDF fulltext, annotations, and Zotero notes which could contain hidden instructions. 1. Ingestion points: item pdf, item fulltext, item annotation list, and item note search in SKILL.md. 2. Boundary markers: The skill instructs the agent to summarize results before showing raw data, though no formal delimiters are specified. 3. Capability inventory: Shell command execution (zot), file writing (item download), and network operations (Zotero Web API). 4. Sanitization: No explicit sanitization or filtering of PDF/note content is documented.
- [COMMAND_EXECUTION]: The skill instructs the agent to execute the zot CLI and cargo build tools for library management. Evidence: Command patterns such as 'zot --json' and fallback 'cargo run -q -p zot-cli' in SKILL.md.
- [DATA_EXFILTRATION]: The skill handles sensitive credentials and reads local SQLite databases. Evidence: Access to ZOT_API_KEY and zotero.sqlite for library queries. The skill explicitly forbids including keys in output or logs.
- [EXTERNAL_DOWNLOADS]: The skill references automated downloads for the Pdfium library and fetches reports from the external Scite service. Evidence: Mentions of ZOT_PDFIUM_CACHE_DIR and item scite report in SKILL.md.
Audit Metadata