Superagent (@superagent_ai) 's Twitter Profile
Superagent

@superagent_ai

Agent In The Loop Superagent.sh (@ycombinator W24)

ID: 1663839457101897730

linkhttps://www.superagent.sh calendar_today31-05-2023 09:27:17

261 Tweet

1,1K Takipçi

18 Takip Edilen

homanp (@pelaseyed) 's Twitter Profile Photo

Claude Code has some interesting system prompts. The <system-reminder> is especially interesting. As well as passing the current git status and other system information to the system prompt.

Claude Code has some interesting system prompts. The &lt;system-reminder&gt; is especially interesting. As well as passing the current git status and other system information to the system prompt.
homanp (@pelaseyed) 's Twitter Profile Photo

I really like the built in VibeKit dashboard we shipped yesterday, built with Next.js and shadcn. Gives me a real time view of what everyone is working on with one cmd 🤯

I really like the built in VibeKit dashboard we shipped yesterday, built with <a href="/nextjs/">Next.js</a>  and <a href="/shadcn/">shadcn</a>.

Gives me a real time view of what everyone is working on with one cmd 🤯
homanp (@pelaseyed) 's Twitter Profile Photo

Excited to bring VibeKit CLI support for OpenAI codex ⚡️ npm i -g vibekit ✅ Realtime redaction of API keys + sensitive data ✅ Observability dashboard (runs, file changes) ✅ Sandboxing your local environment

homanp (@pelaseyed) 's Twitter Profile Photo

I used to hate wiring CI/CD pipelines just to deploy code to AWS or GCP. Always defaulted to “easy” platforms like Vercel or Railway… but paid the price in $$$. Now I can just vibe-code my own pipeline straight to bare metal. Faster, cheaper, and way more satisfying. 1/ From

I used to hate wiring CI/CD pipelines just to deploy code to AWS or GCP.

Always defaulted to “easy” platforms like Vercel or Railway… but paid the price in $$$.

Now I can just vibe-code my own pipeline straight to bare metal.

Faster, cheaper, and way more satisfying.

1/ From
VibeCoder, MD (@vibecodermcswag) 's Twitter Profile Photo

homanp This is actually highly useful So to break it down, you just pipeline the AWS routine into GH actions, so the AI agent / Claude code doesn’t fuck up spinning the instance / task / balance loader / image / all that stuff in the right order. Is this right?

homanp (@pelaseyed) 's Twitter Profile Photo

Tiny checklist before you ship your next AI app: ✅ Redact sensitive data in+out ✅ Block prompt injection & jailbreaking ✅ Scan for PII leaks ✅ Log everything, review often ✅ Limit agent permissions Don't trust best intentions. Trust the guardrails you *actually* test.

Fondo (@tryfondo) 's Twitter Profile Photo

🚀 Superagent launched VibeKit! The Safety Layer for Your Coding Agent "Open-source sandboxing and observability for Claude Code, Gemini CLI, et al." 🌐 fondo.ai/3HAVhnk Congrats Alan Zabihi homanp!!

homanp (@pelaseyed) 's Twitter Profile Photo

You can now use Cursor CLI and Opencode without wrecking anything! ✅ Stops destructive commands from trashing your local machine ✅ Redacts secrets and API keys in real time ✅ Lets you actually see what’s going on npm i -g vibekit

You can now use Cursor CLI and Opencode without wrecking anything!

✅ Stops destructive commands from trashing your local machine
✅ Redacts secrets and API keys in real time
✅ Lets you actually see what’s going on

npm i -g vibekit
VibeKit (@vibekit_sh) 's Twitter Profile Photo

🚀 Excited to announce that VibeKit CLI now supports Cursor CLI and OpenCode! This was by popular demand. Use whichever agent you prefer with the same safety layer.

🚀 Excited to announce that VibeKit CLI now supports Cursor CLI and OpenCode!

This was by popular demand. Use whichever agent you prefer with the same safety layer.
homanp (@pelaseyed) 's Twitter Profile Photo

Gemma 3 (270M) looks like a big step for compact AI chips. With lower latency and training time/cost you can now basically train models in realtime. Imagine deploying agents to the edge.

Gemma 3 (270M) looks like a big step for compact AI chips. With lower latency and training time/cost you can now basically train models in realtime. Imagine deploying agents to the edge.
homanp (@pelaseyed) 's Twitter Profile Photo

This simple prompt injection got read/write access to 1M repositories. Simple static analysis tools won't spot it. Your LLM observability layer that promises "security" might flag it, but by then the damage is done. In order truly secure model inputs/outputs and actions you

This simple prompt injection got read/write access to 1M repositories. 

Simple static analysis tools won't spot it. 

Your LLM observability layer that promises "security" might flag it, but by then the damage is done. 

In order truly secure model inputs/outputs and actions you