integrate

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (high risk: 0.80). The skill explicitly calls docs-mcp-server:explore_product and then docs-mcp-server:doc_fetch_tool to fetch external documentation pages (the "md content link" URLs in $DOC_MAP) and the agent must read and interpret those fetched docs in Phase 3 and later phases to drive codegen, config, webhooks, and tests, so third‑party doc content can materially influence actions.

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 0.90). The skill calls docs-mcp-server:doc_fetch_tool({ url: "<md content link from $DOC_MAP>" }) at runtime to fetch Markdown documentation pages (the "md content link" URLs returned by docs-mcp-server:explore_product), and those fetched pages are used as authoritative input that directly controls prompts, generated code, and validation logic — therefore these runtime-fetched doc URLs are a high-confidence risky external dependency.

MEDIUM W009: Direct money access capability detected (payment gateways, crypto, banking).

• Direct money access detected (high risk: 1.00). The skill is explicitly built to integrate with a payment gateway (Juspay). It contains product-specific MCP calls and actions that provision API keys (juspay_create_api_key), query and update merchant/webhook/general settings (juspay_get_merchant_details, juspay_get_general_settings, juspay_get_webhook_settings, juspay_update_webhook_settings, juspay_update_general_settings), and orchestration for creating payment sessions and running end-to-end payment tests (session creation scripts, webhook event posting, test cards/UPI flows). These are specific, non-generic financial operations targeting a payment processor and enabling live payment flows and configuration — not generic browser/HTTP tooling. Therefore it grants Direct Financial Execution Authority.

MEDIUM W021: Hidden or invisible Unicode characters detected (potential obfuscation or prompt injection).

• Hidden Unicode characters detected (1 type(s) found)