statement-auditor
Pass
Audited by Gen Agent Trust Hub on Sep 1, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process external financial documents (PDF, Excel, CSV) which are considered untrusted data sources.
- Ingestion points: Data enters the agent's context through tools like
statement-reader,document-parser, andocr-and-documentsused to extract content from LP reports. - Boundary markers: The instructions include a 'Report untrusted' principle and suggest the use of a 'Security Isolation Area (Sandbox)', though it lacks specific text-based delimiters for the extracted content.
- Capability inventory: The skill utilizes file parsing capabilities and interacts with external financial data via Tushare MCP tools (
mcp_tushare_fund_nav,mcp_tushare_daily_basic, etc.). - Sanitization: The skill emphasizes numeric reconciliation and cross-validation against a trusted baseline (NAV data), which naturally filters malicious instructional content intended to influence calculations, though it does not explicitly sanitize text fields.
Audit Metadata