I build reliable AI agents and the tools around them.

I build small, open-source tools for the parts of agent engineering that are easy to ignore and expensive to get wrong: context, evaluation, safety, and developer experience.

A local, zero-API-key audit for the instructions, skills, and MCP configuration your coding agents actually consume.

It maps context cost, catches duplicated instructions, flags risky configuration, and produces a scorecard that can run in CI.

• Ship runnable tools, not AI theater.
• Measure context before adding more context.
• Keep the default path local and inspectable.
• Treat evals and failure cases as product features.
• Support the agent stack people already use.

If you are working on agent infrastructure, open an issue or start a discussion.