matechat-vue

Warn

Audited by Gen Agent Trust Hub on Sep 15, 2026

Risk Level: MEDIUMDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [DYNAMIC_EXECUTION]: The demo implementation for the MarkDownCard component in references/components/markDownCard/demo.md uses the new Function() constructor to parse configuration strings. This pattern involves executing strings as JavaScript code, which poses a significant security risk if the data source (such as an LLM response) is untrusted or susceptible to injection attacks.\n- [INDIRECT_PROMPT_INJECTION]: The components provided by this skill are intended to ingest and display content from external AI models, creating a surface for indirect prompt injection. \n
  • Ingestion points: McBubble (content prop), McMarkdownCard (content prop) in references/components/bubble/api.md and references/components/markDownCard/api.md.\n
  • Boundary markers: The documentation suggests developers implement their own handling, though no specific boundary markers are enforced by the components themselves.\n
  • Capability inventory: Includes file upload triggers via McAttachment and network request handling for various components documented in references/components/attachment/api.md.\n
  • Sanitization: The skill documents support for custom XSS filtering rules and uses standard markdown parsing libraries as seen in references/components/markDownCard/demo.md.\n- [EXTERNAL_DOWNLOADS]: The documentation guides users to install various dependencies from official package registries, including @matechat/core, openai, and mermaid. It also fetches the ECharts library from a public CDN (cdnjs) within the demo code in references/components/markDownCard/demo.md.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Sep 15, 2026, 03:49 PM
Security Audit — agent-trust-hub — matechat-vue