month-end-prep

Pass

Audited by Gen Agent Trust Hub on Jul 3, 2026

Risk Level: SAFEDATA_EXFILTRATIONCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [DATA_EXFILTRATION]: The skill is designed to access and process sensitive business financial information, including Profit and Loss (P&L) statements, transaction registries, and bank/acquiring reports from services like 1C, MoySklad, YooKassa, and Tinkoff. This access is necessary for the skill's primary function of financial reconciliation.
  • [COMMAND_EXECUTION]: The skill performs local file system operations, such as scanning the ~/Documents/Первичка directory for source documents and exporting generated XLSX and PDF reports to the user's Desktop. It also references the use of Python libraries xlsxwriter and reportlab to facilitate this file generation.
  • [PROMPT_INJECTION]: The skill possesses a surface for indirect prompt injection as it ingests and summarizes untrusted external data from financial records and local file metadata.
  • Ingestion points: Transaction descriptions from accounting APIs, banking statements, and filenames/metadata from local files in ~/Documents/Первичка.
  • Boundary markers: The workflow implements a mandatory approval gate at Step 6, requiring the agent to present reconciliation flags to the user and obtain explicit confirmation before generating final summaries or files.
  • Capability inventory: The agent has the ability to read and write to the local file system, access financial data via connected tools, and generate complex document formats.
  • Sanitization: No explicit sanitization or escaping mechanisms are described for transaction descriptions before they are processed into the plain-language P&L narrative in Step 7.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 3, 2026, 06:56 PM
Security Audit — agent-trust-hub — month-end-prep