lekkalaharsha/defending-code-reference-harness — reverse-engineered prompt

Build me a reference security workflow I can open in Claude Code to find and fix code vulnerabilities.

I want a simple guided experience where I can run slash commands like /quickstart, /threat model, /vuln scan, /triage, /patch, and /customize. It should help me start with a threat model, do a static scan, review and rank findings, and then generate candidate fixes. I also want an autonomous pipeline version that goes through recon, finding bugs, verifying them, writing a report, and proposing patches.

Please make it clear that this is a reference harness, not a polished product. The autonomous part should be set up for C and C plus plus memory bug hunting, with execution based verification, and it needs to run inside a proper sandbox for safety. The read only style skills should stay safe for normal interactive use, while anything that runs target code should refuse to run unless the sandbox is set up.

Include a small canary target and docs so someone can try it fast, then customize it for their own stack. Look up current docs online if you need to.