difftool
Pass
Audited by Gen Agent Trust Hub on Oct 29, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill provides legitimate instructions for visual regression testing using the vendor's tooling. It correctly identifies sensitive information like API keys and S3 configurations and recommends using environment variables or repository secrets for their storage.
- [EXTERNAL_DOWNLOADS]: The CLI usage instructions involve cloning the repository from GitHub (
microlinkhq/difftool) and installing dependencies viapnpm. These resources originate from the skill author's verified infrastructure. - [COMMAND_EXECUTION]: The skill describes executing a local Node.js script (
node packages/cli/bin/cli.js) after cloning the repository. This is a standard workflow for running a development tool that is not published to a package registry. - [INDIRECT_PROMPT_INJECTION]: The tool's primary function is to process external URLs to capture screenshots.
- Ingestion points: External URLs provided as arguments to the
--baseand--headflags (SKILL.md). - Boundary markers: None identified in the tool's command-line interface instructions.
- Capability inventory: The tool executes a Node.js CLI that performs network requests to the Microlink API and writes screenshot files to a local directory or uploads them to an S3 bucket.
- Sanitization: Not applicable, as the tool's output consists of image diffs and a structured JSON summary rather than natural language instructions processed by the agent.
Audit Metadata