←
Kyle Avery1,166 posts
Kyle Avery avatar

Kyle Avery

@kyleavery

coors reaper

◷ Joined July 2017▧ 147 Photos and videos
Kyle Avery avatar
Kyle Avery@kyleavery·Nov 17

Usually I dismiss concerns with LLM guardrails/refusals (for maldev, at least). Try harder. Using the CVP Red Team/Daybreak Red models sure saves time though lol

Kyle Avery avatar
Kyle Avery@kyleavery·Nov 12

Regardless, it’s great to see open-source examples of both

I get the appeal of putting an agent harness in your C2 interface, but I don't think it makes sense. It (usually) means I can't reuse my ChatGPT/Claude subscription, and an agent that runs on the server won't have the tools I'd run over SOCKS or the local BOFs I want it to extend/debug. Beyond this, building a good harness is difficult. Just look at the Codex or OMP repos... many decisions impact model performance (prompt caching, compaction, tool output truncation, etc.).

Kyle Avery avatar
Kyle Avery@kyleavery·Nov 11

I get the appeal of putting an agent harness in your C2 interface, but I don't think it makes sense. It (usually) means I can't reuse my ChatGPT/Claude subscription, and an agent that runs on the server won't have the tools I'd run over SOCKS or the local BOFs I want it to extend/debug. Beyond this, building a good harness is difficult. Just look at the Codex or OMP repos... many decisions impact model performance (prompt caching, compaction, tool output truncation, etc.).

Kyle Avery avatar
Kyle Avery@kyleavery·Oct 30

“we don’t use your data for training 🤓” ok that’s cool but i don’t want to use your model

Kyle Avery avatar
Kyle Avery@kyleavery·Oct 30

if you have issues with the new Cobalt Strike MCP+Skill in Codex/Claude Code, please let me know! of course it works best with Daybreak/CVP

Kyle Avery avatar
Kyle Avery@kyleavery·Oct 10

anybody know how to start a security conference? i need an excuse to bring everyone to Texas

dreadnode avatar
⟳ Kyle Avery retweeted
dreadnode@dreadnode·Oct 6

🧵 [1/3] Agents can hack. What's holding them back is trust: will they stay in scope? Today we're releasing ScopeBench, a methodology and community benchmark that evaluates how well agents adhere to scope in real offensive security workflows across web, Windows Active Directory, and cloud.

Post attachment