visual-ralph

Pass

Audited by Gen Agent Trust Hub on May 14, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill directs the agent to execute local project commands, including package manager scripts (npm, yarn, pnpm), build processes, and linting/testing suites. It also requires the execution of a screenshot command to capture the current state of the UI for automated visual comparison.
  • [EXTERNAL_DOWNLOADS]: The skill facilitates the capture of visual references from external live URL targets. This involves accessing remote web content to establish a baseline for UI cloning or migration tasks.
  • [PROMPT_INJECTION]: The skill is designed to ingest and process content from external URLs and user descriptions, which represents an indirect prompt injection surface.
  • Ingestion points: External URLs and user-supplied descriptions in SKILL.md.
  • Boundary markers: The workflow mentions documentation of viewports and interaction notes, but lacks explicit boundary markers or instructions for the agent to ignore instructions embedded within the processed external content.
  • Capability inventory: The skill orchestrates file system modifications via Ralph, command execution for builds/tests, and network access for URL capture.
  • Sanitization: There is no mention of sanitization or filtering for content retrieved from external URLs.
Audit Metadata
Risk Level
SAFE
Analyzed
May 14, 2026, 03:59 AM
Security Audit — agent-trust-hub — visual-ralph