univer-node-backend
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill documents features for processing spreadsheets from external sources, which represents a standard ingestion surface for indirect prompt injection.\n
- Ingestion points: Data is processed from databases via
db.query, local JSON files viafs.readFileSync, and remote XLSX files viaapi.importXLSXToUnitIdAsync.\n - Boundary markers: Code examples show direct processing of data without specific delimiters or warnings to the agent to ignore instructions embedded in cell content.\n
- Capability inventory: The integration involves file system write access (
fs.writeFileSync) and the ability to register custom formula functions, which are core library features but could be misused by malicious input data.\n - Sanitization: As a technical integration guide, the examples focus on functionality and do not explicitly demonstrate spreadsheet content validation or sanitization.\n\n- [DYNAMIC_EXECUTION]: The skill describes the use of child processes for offloading formula execution to worker threads.\n
- Evidence: The usage of
UniverRPCNodeMainPlugininSKILL.mdandreferences/node-common-tasks.mdinvolvesnode:child_process.fork()to spawn formula workers based on locally provided script paths (worker.js).\n - Assessment: This is a routine architectural pattern for performance in Node.js applications and does not present an unusual security risk in the context of this library.
Audit Metadata