extract-bank-statement

Pass

Audited by Gen Agent Trust Hub on Jul 29, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill executes shell commands to probe PDF metadata and run local Python scripts (scripts/extract_maybank_islamic_pdf.py and scripts/normalize_bank_csv.py) for data extraction.
  • [PROMPT_INJECTION]: The skill ingests untrusted external data from bank statements, which could theoretically contain malicious instructions. 1. Ingestion points: user-provided PDF and CSV files. 2. Boundary markers: no specific delimiters are defined to isolate statement text from the agent's logic. 3. Capability inventory: the skill executes Python scripts via the shell and writes output files to the local filesystem. 4. Sanitization: it uses mathematical balance proofs and a --fail-on-error flag to ensure data integrity during processing.
  • [DATA_EXPOSURE]: The skill is designed to handle and transform highly sensitive financial information, such as transaction history and account balances, into machine-readable JSON and Excel formats.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 29, 2026, 10:04 AM
Security Audit — agent-trust-hub — extract-bank-statement