visual-feedback-loop
Pass
Audited by Gen Agent Trust Hub on Sep 17, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The reference implementation for Next.js uses
execSyncto retrieve Git commit information and repository status. These commands are hardcoded and intended for metadata tracking in a development environment. - [COMMAND_EXECUTION]: The skill provides instructions for the agent to use the
magick(ImageMagick) CLI tool to perform visual regression analysis by comparing image files stored in the local.screenshots/directory. - [EXTERNAL_DOWNLOADS]: The documentation mentions the installation of
webmcp-kitandzodvia npm to enable WebMCP tool registration, which allows browser-integrated agents to invoke the screenshot functionality directly. - [INDIRECT_PROMPT_INJECTION]: The skill directs the agent to read generated image files (
.screenshots/latest.webp). Because the agent may use a vision model to interpret the contents of the rendered UI, there is an inherent surface for indirect prompt injection if the UI contains malicious text, although the risk is limited to the local development context.
Audit Metadata