Loading video...

Video Failed to Load

Go Home

Frontier LLM models (especially Opus) are very good at finding vulnerabilities in codebases. With the right skills and a sub-agent architecture, they can outperform any traditional SAST tool. I collected the skills I use in this repository. Load them into any AI coding IDE such as Claude Code, Codex,...

95,266 views • 7 months ago •via X (Twitter)

9 Comments

Aziz's profile picture
Aziz7 months ago

thanks man 🤝 let's see how it goes

BannMe's profile picture
BannMe7 months ago

the flip side of this keeps me up at night. the same models writing our code are better at finding vulns than the tools we use to catch them.

Alican Selçuk's profile picture
Alican Selçuk7 months ago

Şuan tam ihtiyacım olan şeydi. Teşekkürler, harikasın koca yürekli siber güvenlikçi :)

emrec's profile picture
emrec7 months ago

👏👏

Ozgur Ozkan's profile picture
Ozgur Ozkan7 months ago

Cool for defense but what about active defense? @audn_ai AI Active Defense

sonic dev's profile picture
sonic dev7 months ago

that’s what we need. thanks!

Boyuan (Nemo) Chen's profile picture
Boyuan (Nemo) Chen7 months ago

Single-pass scanning consistently misses the context-dependent stuff, which is why the sub-agent split matters more than the model choice here. Curious whether you decompose by vuln class or by analysis stage (detection vs. exploitability reasoning)? Those two seem to need fundamentally different prompting.

Utku Sen 🇬🇧's profile picture
Utku Sen 🇬🇧7 months ago

there are subagent splits inside SKILLs. have you checked the repo?

Gagan | Claude + AWS's profile picture
Gagan | Claude + AWS7 months ago

Opus with a multi-agent security setup is genuinely scary good at this. The pattern recognition across large codebases + ability to chain sub-agents for deep dives outclasses most traditional static analysis tools.

Related Videos