usestrix/strix ? reverse-engineered prompt

I want to build an open-source security tool that uses AI agents to automatically find vulnerabilities in my apps. Imagine it as a team of AI hackers that can dynamically run an application in a sandboxed environment, like Docker, to find and validate real exploits. The main goal here is to get actual proof-of-concepts and avoid the false positives that you see with typical static analysis tools.

It should be a developer-first CLI tool. I need to be able to just point it at a local codebase, a GitHub repo, or a live URL to kick off a scan. The tool should then generate actionable reports for any issues it discovers, and maybe even suggest fixes. It should be configurable with an LLM API key from major providers like OpenAI or Anthropic. A critical feature is seamless integration into a CI/CD pipeline, ideally with a simple GitHub Action to automatically scan new pull requests before they get merged.