dingtalk-aicard
Pass
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONNO_CODE
Full Analysis
- [COMMAND_EXECUTION]: The skill provides instructions for the agent to execute the
dwscommand-line utility to perform actions such as protocol lookup (explain), structural validation (lint), and sending card previews to the current user (preview). - [INDIRECT_PROMPT_INJECTION]: The skill processes user-provided requirements and images to generate A2UI JSON card definitions. These definitions are subsequently handled by the
dwstool, creating a potential attack surface for indirect prompt injection via the card's content. - Ingestion points: User requirements, reference images, and generated or edited A2UI JSON files.
- Boundary markers: The protocol uses structured JSON messages to isolate different card components and data model fields.
- Capability inventory: The skill uses the
dwsbinary for card validation and previewing. - Sanitization: The skill explicitly requires running the
dws aicard lintcommand with preflight checks to validate the card's structure and resource encoding before any delivery. - [NO_CODE]: The skill does not include any executable scripts (such as Python or JavaScript); it consists entirely of Markdown instructions, JSON schemas, and example protocol messages.
Audit Metadata