match
Pass
Audited by Gen Agent Trust Hub on Nov 5, 2026
Risk Level: SAFEDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [DYNAMIC_EXECUTION]: The skill instructs the agent to create a JavaScript file (
compare.js) and a comparison HTML page (compare.html), then open the dashboard to view results. - Evidence:
SKILL.mdinstructions for step 5 specify: "Copy compare.html to .match//index.html... Write compare.js beside it in the shape its header comment shows... then open it." - [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted content from external reference sources, including live web pages and design files.
- Ingestion points:
SKILL.mdlists reference sources such as live pages, installed apps, and design files. - Boundary markers: The instructions lack explicit delimiters or safety warnings to ignore instructions that might be embedded in the UI content being matched.
- Capability inventory: The skill can write files to the local file system and open them in a browser.
- Sanitization: There is no requirement for the agent to sanitize the data extracted from the source before writing it to the generated script file.
- [EXTERNAL_DOWNLOADS]: The generated comparison tool downloads assets from a well-known service during rendering.
- Evidence:
compare.htmlincludes<link>tags forfonts.googleapis.comandfonts.gstatic.comto load the Inter font family.
Audit Metadata