PortfolioSyncing

Warn

Audited by Gen Agent Trust Hub on Feb 16, 2026

Risk Level: MEDIUMDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
  • Data Exposure & Exfiltration (MEDIUM): The skill reads highly sensitive financial files including Portfolio_Positions_*.csv and Balances_for_Account_*.csv from the local notebooks/updates/ directory and transmits this data to a remote Google Sheets spreadsheet via tool calls.
  • Indirect Prompt Injection (MEDIUM): The workflow is vulnerable to indirect prompt injection via manipulated CSV files. 1. Ingestion points: Local CSV files in notebooks/updates/. 2. Boundary markers: Absent; data is extracted and interpolated directly into tool parameters. 3. Capability inventory: mcp__gdrive__sheets update operations. 4. Sanitization: Absent; the agent processes CSV column data without validation.
  • Safe Practices (INFO): The workflow implements 'Safety Checks' in Step 5, requiring explicit user confirmation before proceeding with updates that exceed specific thresholds, such as ticker counts or percentage changes in cost basis.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Feb 16, 2026, 01:35 AM