The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill recommends installing the Node.js package israeli-bank-scrapers and specifically suggests a fork @sergienko4/israeli-bank-scrapers to bypass Cloudflare bot detection. The use of individual user forks for financial data processing introduces a supply chain risk.
[COMMAND_EXECUTION]: The instructions guide the user to enable the Execute Command node in n8n 2.0 using the N8N_COMMUNITY_PACKAGES_ALLOW_TOOL_USAGE environment variable. While presented as a migration path for legacy workflows, this configuration significantly expands the attack surface by allowing shell command execution from within the automation platform.
[PROMPT_INJECTION]: The skill identifies an indirect prompt injection surface in the bank transaction categorization workflow in SKILL.md (Step 7).
Ingestion points: External transaction descriptions fetched via israeli-bank-scrapers and payment data received through Webhook nodes.
Boundary markers: Absent in the provided prompt examples; external data is interpolated directly into the task instructions for the AI Agent.
Capability inventory: The skill uses Code nodes for arbitrary JavaScript execution and HTTP Request nodes for network operations, which could be triggered by instructions embedded in transaction data.
Sanitization: No sanitization or filtering is applied to the Hebrew transaction descriptions before they are processed by the LLM.

n8n-hebrew-workflows