Chirag

[Chirag @ChiragCX]

@victormustar Open source?

[Chirag @ChiragCX]

@OpenAINewsroom @ona_hq So enterprise grade open claw .. smart

[Chirag @ChiragCX]

@altryne what was your prompt?

[Chirag @ChiragCX]

@TinoWening yes DAGs! I have been thinking about this as well. Do you have any examples of how you have put this together?

[Chirag @ChiragCX]

@OpenAI How is this difffernt than the browser use tool?

[Chirag @ChiragCX]

@svpino I don’t think you understand how harness works. Claude and codex, etc post train on tools call, this just a generic harness , just use deep agents

[Chirag @ChiragCX]

@dbreunig whats wrong with that, can't skills just be the new form of documentation?

[Chirag @ChiragCX]

@LLMJunky Yeah I created a post skill feedback skill (mouthful) I attached this to all skills at the end to have it reflect on itself an any assumptions that it has to take because the skill lacked the right context and save the feedback to a table for me to review ,

[Chirag @ChiragCX]

@hwchase17 A multi tenant claw/hermes like gateway that uses deepagent as the harness

[Chirag @ChiragCX]

@bllchmbrs opus did a manager view on you, review said "skill issue.."

[Chirag @ChiragCX]

@YokoHyper @hwchase17 deepagents is actually pretty amazing. I build a gateway like openclaw and deepagents is my harness at my company. its amainzg!

[Chirag @ChiragCX]

@matt_ambrogi Just build it, it’s a loop

[Chirag @ChiragCX]

@MikeBirdTech @pmarca What do you mean by “ntegrate a request router”

[Chirag @ChiragCX]

@anthonyronning @LakshyAAAgrawal @MaximeRivest Tell us more!

[Chirag @ChiragCX]

@sawyerhood amazing!! in the previous iteration you had extention mode to connect to logged in profiles. How does that work now?

[Chirag @ChiragCX]

@ocbieuvang @badlogicgames @LakshyAAAgrawal open source?

[Chirag @ChiragCX]

@LLMJunky Unsloth! https://t.co/JaiH0kbUCr

[Chirag @ChiragCX]

curious if ppl are running their skills eval in the CI/CD pipeline, using a github action whenever a skill is trying to be pushed?

[Chirag @ChiragCX]

@OpenAI Do I just use autoresearch?

[Chirag @ChiragCX]

@VladBarash also this point if your going to do all this work, just do your own RLVR with something like verifiers by @willccbb

[Chirag @ChiragCX]

@VladBarash yeah that is exactly it. The hardest part is how to do you come up with the rubric. You have to ask yourself "what is your goal" if its agent optimization with objective measures ie a agentic benchmark eval but just for your metrics. I would use autoresearch