Security-audit a codebase with a skill where a second agent tries to disprove every finding
Cloudflare open-sourced its security-audit skill. Point a coding agent at a codebase and say "security audit this codebase": it maps trust boundaries, hunts attack types with parallel agents, has a fresh agent try to prove each bug wrong, saves findings as schema-checked JSON, verifies them against the source and writes a report.
Source
@Ldataengineer on X · 2026-10-22
Written in our own words. Credit to the original author. Link checked 2026-10-23.
What they did
Cloudflare open-sourced its security-audit skill. Point a coding agent at a codebase and say "security audit this codebase": it maps trust boundaries, hunts attack types with parallel agents, has a fresh agent try to prove each bug wrong, saves findings as schema-checked JSON, verifies them against the source and writes a report.
Result (as claimed)
Separating finding from checking cuts the false alarms that waste security teams' hours.
Tools they used
Cost: Claude Code: from $17/mo.
Set it up
Pick a contained repo or task
Start on a branch or side project where mistakes are cheap.
Install and configure
Set up Claude Code, Cloudflare security-audit skill. Read what it can run on your machine and limit permissions where possible.
Run it on a real task
Give it a well-scoped issue and let it work. Review the diff like any pull request.
Add guardrails
Keep tests and CI in the loop, and never let an agent push to main or touch production secrets unattended.
Copy the full setup prompt
Paste into your agent and it will walk you through the setup, asking before anything that sends, posts or spends.
You are helping me set up this AI use case end to end in Claude Code. Goal: Security-audit a codebase with a skill where a second agent tries to disprove every finding What the original author did: Cloudflare open-sourced its security-audit skill. Point a coding agent at a codebase and say "security audit this codebase": it maps trust boundaries, hunts attack types with parallel agents, has a fresh agent try to prove each bug wrong, saves findings as schema-checked JSON, verifies them against the source and writes a report. Reported result (as claimed by the author): Separating finding from checking cuts the false alarms that waste security teams' hours. Source: https://x.com/Ldataengineer/status/2113233908238315830 Tools they used: Claude Code, Cloudflare security-audit skill Category: Dev & ops Please: 1. Ask me 3 short questions about my situation (accounts I have, budget, what I want the output to look like). 2. Propose a step-by-step plan using these tools or cheaper alternatives, with the time and cost of each step. Use official pricing pages; say "check pricing" if unsure. 3. Then do the setup with me one step at a time, waiting for my OK before anything that sends messages, posts publicly, spends money or deletes data. 4. Finish with a checklist to run it again next time.
You are helping me set up this AI use case end to end in Codex. Goal: Security-audit a codebase with a skill where a second agent tries to disprove every finding What the original author did: Cloudflare open-sourced its security-audit skill. Point a coding agent at a codebase and say "security audit this codebase": it maps trust boundaries, hunts attack types with parallel agents, has a fresh agent try to prove each bug wrong, saves findings as schema-checked JSON, verifies them against the source and writes a report. Reported result (as claimed by the author): Separating finding from checking cuts the false alarms that waste security teams' hours. Source: https://x.com/Ldataengineer/status/2113233908238315830 Tools they used: Claude Code, Cloudflare security-audit skill Category: Dev & ops Please: 1. Ask me 3 short questions about my situation (accounts I have, budget, what I want the output to look like). 2. Propose a step-by-step plan using these tools or cheaper alternatives, with the time and cost of each step. Use official pricing pages; say "check pricing" if unsure. 3. Then do the setup with me one step at a time, waiting for my OK before anything that sends messages, posts publicly, spends money or deletes data. 4. Finish with a checklist to run it again next time.
You are helping me set up this AI use case end to end in Grok Bot. Goal: Security-audit a codebase with a skill where a second agent tries to disprove every finding What the original author did: Cloudflare open-sourced its security-audit skill. Point a coding agent at a codebase and say "security audit this codebase": it maps trust boundaries, hunts attack types with parallel agents, has a fresh agent try to prove each bug wrong, saves findings as schema-checked JSON, verifies them against the source and writes a report. Reported result (as claimed by the author): Separating finding from checking cuts the false alarms that waste security teams' hours. Source: https://x.com/Ldataengineer/status/2113233908238315830 Tools they used: Claude Code, Cloudflare security-audit skill Category: Dev & ops Please: 1. Ask me 3 short questions about my situation (accounts I have, budget, what I want the output to look like). 2. Propose a step-by-step plan using these tools or cheaper alternatives, with the time and cost of each step. Use official pricing pages; say "check pricing" if unsure. 3. Then do the setup with me one step at a time, waiting for my OK before anything that sends messages, posts publicly, spends money or deletes data. 4. Finish with a checklist to run it again next time.
The Daily 5
Get five new use cases tomorrow morning
Free. Real sources. Unsubscribe any time.
Guides for this use case
How to run multiple coding agents together · How to use Claude Code skills and plugins · How to secure vibe-coded apps and agents · How to vibe code an app with Claude Code
Related use cases
Scope an AI security audit to one folder so it doesn't burn your tokens
Cloudflare's open-source, MIT-licensed security audit skill installs with npx skills add and runs when you prompt "security audit this codebase". Because several agents double-check each finding it is slow and costly on big monorepos, so scope it to specific directories.
Track security news against your exact stack with Claude Code and Grok Bot
A non-security founder had Claude Code list every platform, package, version and third-party dependency in his product, then gave that inventory to Grok Bot to keep stored and check against new vulnerabilities.
Get most of an enterprise AI security scanner for free with open-source repos
With Claude Security limited to Enterprise, a roundup lists free repos that run on Claude Code: airsec for local multi-pass audits, paranoid's /hack-me that attacks your running app and proves each bug with a real request, security-harness with SARIF/JSON/PDF reports, claude-security-scan for a read-only sweep, prod-readiness with a SHIP/FIX/HOLD verdict, and jev-guard that scores every tool call.
Run one chat where Claude plans, another model writes and Codex reviews
Shikidan is one chat that runs your coding agents: Claude plans, GLM writes the code, and Codex reviews it before it lands on your branch, so a different AI checks the work.