retirement-syncing

Pass

Audited by Gen Agent Trust Hub on Feb 17, 2026

Risk Level: SAFENO_CODE
Full Analysis
  • Data Exposure (SAFE): The skill reads from local financial CSVs and a user profile file (fin-guru/data/user-profile.yaml). This access is strictly contained within the agent's local environment and is necessary for identifying the target spreadsheet and processing account data as intended.- Indirect Prompt Injection (SAFE): The skill processes external CSV files provided by the user.
  • Ingestion points: notebooks/retirement-accounts/OfxDownload.csv, notebooks/retirement-accounts/OfxDownload (1).csv, and notebooks/retirement-accounts/Portfolio_Positions_*.csv.
  • Boundary markers: The skill relies on structured column mapping (Symbol, Shares/Quantity) rather than natural language parsing of the file content, which limits the influence of embedded text.
  • Capability inventory: File system read access and targeted Google Sheets write access (mcp__gdrive__sheets).
  • Sanitization: The implementation uses numeric type casting (float()) and formatting (toFixed) on the ingested data, effectively preventing the execution of arbitrary text instructions embedded in the quantity fields.
Audit Metadata
Risk Level
SAFE
Analyzed
Feb 17, 2026, 06:23 PM