mcp-lark

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 1.00). The prompt instructs embedding MCP tokens directly into command-line URLs/arguments (e.g., open.larksuite.com/mcp/stream/) and asks the agent to fetch or request those tokens and place them into commands, which requires the LLM to handle and output secret values verbatim.

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

• Third-party content exposure detected (high risk: 0.80). The SKILL.md explicitly instructs the agent to call MCP endpoints on open.larksuite.com (see LARK_MCP_SERVERS and npx mcporter calls) to manage chats, emails and cloud documents—i.e., user-generated, public third-party content that the agent will fetch and interpret as part of its workflow, enabling indirect prompt injection.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill invokes runtime endpoints such as "open.larksuite.com/mcp/stream/" via commands like "npx -y mcporter …" (which fetches and executes the mcporter package from the npm registry) so remote MCP tool schemas and code are fetched at runtime and can directly control agent tooling/instructions or execute code.