Projects

Tools I build and maintain for application security and AI agent testing.

Stealth Prompt

A browser-native assistant for authorized AI agent testing. It works against the authenticated chat UI already open in Chrome, so a test runs with real session state instead of a recreated one.

  • Runs against the authenticated UI and account state already open in Chrome.
  • You pick the input, the send control, and the response container, so the model cannot broaden the target on its own.
  • Generates a test message, reads the reply, and continues the same line of investigation.

AI Agent Security Starter

A minimal repo showing how to put guardrails around AI-assisted coding: AGENTS.md, a security skill, hooks, and Python validators for prompts, commands, and sensitive paths.