tear-sheet
Pass
Audited by Gen Agent Trust Hub on Jun 17, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [DYNAMIC_EXECUTION]: The skill utilizes a dynamic approach to document generation by constructing and executing a Node.js script that relies on the 'docx' library. This script-generation process is the intended and necessary mechanism for creating professional Word documents with specific styling and layouts.
- [INDIRECT_PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection due to its core function of ingesting data from external sources. The evidence chain is as follows:
- Ingestion points: Data is retrieved from S&P Global MCP tools across all four audience-specific reference files (e.g., company profiles, earnings transcripts, and business relationships).
- Boundary markers: The intermediate storage files (e.g., /tmp/tear-sheet/financials.csv) do not use explicit delimiters to isolate external data from instructions.
- Capability inventory: The skill has the capability to write to the file system and execute Node.js scripts.
- Sanitization: The skill implements a robust mitigation strategy via Content Quality Rule 11, which explicitly forbids verbatim pasting of external summaries and mandates the analytical rewriting of all narrative sections, thereby reducing the risk of executing embedded malicious instructions.
Audit Metadata