The Agent Skills Directory

[Indirect Prompt Injection] (LOW): The skill creates a vulnerability surface by ingesting untrusted data from external websites via console logs and JavaScript evaluation.
Ingestion points: scripts/chromectl.py console-tail streams logs, and scripts/chromectl.py eval retrieves page state/content.
Boundary markers: No specific delimiters or instructions to ignore embedded commands are present in the workflows.
Capability inventory: The skill can launch processes, navigate to arbitrary URLs, execute JavaScript in the browser context, and capture screenshots.
Sanitization: There is no evidence of sanitization or filtering of data received from the browser before it is processed by the AI agent.
[Command Execution] (LOW): The skill frequently executes a local Python script (scripts/chromectl.py) to manage Chrome instances and interact with the DevTools Protocol. This behavior is aligned with the skill's stated purpose.
[External Downloads] (LOW): The skill documentation indicates it uses uv to manage the aiohttp dependency. While aiohttp is a reputable library, automated package management during runtime presents a minor risk of dependency confusion if not pinned to hashes.

chrome-debug