最近話題の ARTEX ってなんやと思ったら自律的にシステム攻撃に特化したエージェントハーネスなのか…
Daybreak Blue だと取れても防御系だけだし結局攻撃側と同じツールで先に攻撃して攻撃される前に事前に把握して穴塞いどくしかないのか
cskwork/scopeweaver
28 likes · 1 repost · 1,990 views
@NobMiwa てっきりこれのことかと
cskwork/scopeweaver
2 likes · 509 views
ARTEXの英語版クローン上がってた
cskwork/scopeweaver
17 likes · 1,973 views
An LLM multi-agent autonomous penetration-testing system (Go backend + Next.js frontend)
cskwork/scopeweaver
4 likes · 1,211 views
Since some South Korean banks have been successfully attacked using the Chinese Opensource Tool #Artex, the Github repo containing it has been taken down.
Still, there's still a fork (translated into English & Korean) available - you might want to grab it:
cskwork/scopeweaver
@Krongggggg cskwork/scopeweaver 这个是原版。 只是换了个名字。
17 likes · 3 reposts · 757 views
I translated ARTEX, a Chinese multi-agent AI pentest platform, into English and Korean
ARTEX is an open-source autonomous pentest system by Autumn-27 (Go backend, Next.js UI, about 1.5k stars, won Baidu's "agent+" attack/defense challenge). you give it a scoped target, LLM agents work through it, and findings land in a dashboard you can export as markdown or CSV. the catch for most of us was that the UI, agent prompts and docs were Chinese only.
ScopeWeaver is my English/Korean fork. i didn't write the engine. i translated the interface, the built-in agent guidance and 1,865 backend messages, made each task keep the language it was started in so reports come out consistent, and added a GLM-5.3 provider template. license stays AGPL-3.0 with upstream credit.
being upfront since this is a security sub: one upstream test still fails (it fails on untouched upstream too), npm audit shows 14 findings including 1 critical in the inherited dependencies, and there's no docker image yet, source only. you bring your own model API key, so target data goes to whichever provider you configure.
only point it at systems you're authorized to test.
cskwork/scopeweaver
upstream: Autumn-27/ARTEX
curious whether anyone here has run ARTEX or a similar agent on a real engagement.
drafted by my coding agent, reviewed by me
4 points · 1 comment
Mentions
最近話題の ARTEX ってなんやと思ったら自律的にシステム攻撃に特化したエージェントハーネスなのか… Daybreak Blue だと取れても防御系だけだし結局攻撃側と同じツールで先に攻撃して攻撃される前に事前に把握して穴塞いどくしかないのか cskwork/scopeweaver
@NobMiwa てっきりこれのことかと cskwork/scopeweaver
ARTEXの英語版クローン上がってた cskwork/scopeweaver
An LLM multi-agent autonomous penetration-testing system (Go backend + Next.js frontend) cskwork/scopeweaver
Since some South Korean banks have been successfully attacked using the Chinese Opensource Tool #Artex, the Github repo containing it has been taken down. Still, there's still a fork (translated into English & Korean) available - you might want to grab it: cskwork/scopeweaver
@Krongggggg cskwork/scopeweaver 这个是原版。 只是换了个名字。
I translated ARTEX, a Chinese multi-agent AI pentest platform, into English and Korean ARTEX is an open-source autonomous pentest system by Autumn-27 (Go backend, Next.js UI, about 1.5k stars, won Baidu's "agent+" attack/defense challenge). you give it a scoped target, LLM agents work through it, and findings land in a dashboard you can export as markdown or CSV. the catch for most of us was that the UI, agent prompts and docs were Chinese only. ScopeWeaver is my English/Korean fork. i didn't write the engine. i translated the interface, the built-in agent guidance and 1,865 backend messages, made each task keep the language it was started in so reports come out consistent, and added a GLM-5.3 provider template. license stays AGPL-3.0 with upstream credit. being upfront since this is a security sub: one upstream test still fails (it fails on untouched upstream too), npm audit shows 14 findings including 1 critical in the inherited dependencies, and there's no docker image yet, source only. you bring your own model API key, so target data goes to whichever provider you configure. only point it at systems you're authorized to test. cskwork/scopeweaver upstream: Autumn-27/ARTEX curious whether anyone here has run ARTEX or a similar agent on a real engagement. drafted by my coding agent, reviewed by me