zepto-prava-skill

Pass

Audited by Gen Agent Trust Hub on Jul 13, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill utilizes local Node.js utility scripts, zepto-mcp-runner.mjs and zepto-prava-doctor.mjs, which execute system commands such as npx, prava, and codex. These scripts are used to facilitate tool discovery, manage the MCP connection, and diagnose the agent's environment.- [EXTERNAL_DOWNLOADS]: The skill connects to a remote MCP server definition at https://mcp.zepto.co.in/mcp and instructs the agent to install the Prava CLI (@prava-sdk/cli). These external resources are official endpoints for the merchant (Zepto) and the skill's author.- [INDIRECT_PROMPT_INJECTION]: The skill processes data from external sources including product search results, past order history, and saved address labels retrieved via the Zepto MCP server.
  • Ingestion points: Data is retrieved from the Zepto MCP server using tools such as list_saved_addresses, get_past_order_items, and search_products.
  • Boundary markers: The skill instructs the agent to aggregate product items into a single descriptive line for the payment session to improve clarity, though it does not define specific character-level delimiters for the external text.
  • Capability inventory: The skill can execute shell commands through its utility scripts and perform browser automation to complete the Zepto/Juspay payment form.
  • Sanitization: The skill relies on structured JSON-RPC for interaction with the MCP tools, which helps separate data from instructions during tool invocation.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 13, 2026, 08:16 PM
Security Audit — agent-trust-hub — zepto-prava-skill