I made CEO AI: an assistant for talking through tasks and working with AI agents

Description and use

CEO AI is an assistant for task discussion and preventing questionable actions, which also scans external skill and agent instructions to enhance security during installation.

Problem

Developing an AI assistant that can discuss tasks and prevent questionable actions, while ensuring security when installing and connecting external skills and agents.

Outcome

CEO AI enables safer integration of external skills and agents by providing task discussion, safety prompts, and security scans, reducing the risk of executing malicious instructions.

Workflow

• 1. Install CEO AI skill for chosen platform (Codex, Claude Code, Cursor).
• 2. Use the assistant to discuss tasks and receive safety prompts.
• 3. When adding a new skill or agent, run the basic local scan to detect suspicious instructions.
• 4. Optionally run the deeper Docker-based scan for more thorough analysis.
• 5. Review scan results and decide whether to proceed with installation.

Evidence

I made a skill for Codex, Claude Code, and Cursor. I wanted an assistant I could talk through tasks with, and one that would occasionally stop me from doing something questionable when I'm tired.

Source

We don't always check what's written in those instructions. I looked into several scanners, tested them, and added checks during indexing to CEO AI. It examines selected skills and agent definitions, looks for suspicious instructions, and shows what hasn't been checked. It doesn't execute the instructions it's checking.

Source

The basic check runs locally, without Docker or paid API calls. The plugin also includes an optional deeper scan that needs Docker and a prepared scanner image.

Source