The network for creativity
Join 1.25M professional creatives like you
Connect with clients, get discovered, and run your business 100% commission-free
Creatives on Contra have earned over $150M and we are just getting started
Back to feedPost
A great question asked by Huzaifa Tahir on AI Agent Standards... everyone must notice this.
contra.com
Huzaifa Tahir’s Work | Contra
View Huzaifa Tahir’s work on Contra. The Independent-first, commission-free freelance marketplace shaping the future of work.
The OpenSandbox piece is what stands out to me, because safe execution of agent actions matters most when the agent can touch patient-adjacent systems. In healthcare I'd want every tool call scoped to a narrow permission set and logged, so interoperability never quietly widens...
The network for creativity
Join 1.25M professional creatives like you
Connect with clients, get discovered, and run your business 100% commission-free
Creatives on Contra have earned over $150M and we are just getting started
Related posts
Love how each agent has its own identity and personality. Really nice work! 🔥
Work in progress: I built an AI agent to manage my website through Telegram.
A recent conversation with Julian from Native got me thinking about what it means to build AI-native products. I wanted to explore how an agent could take on the routine work of managing my website.
So I built PASH Site Agent using Vercel, Telegram, and Sanity.
Instead of opening a CMS, filling out fields, uploading images, and writing SEO descriptions, I send a message in Telegram.
I describe what I want to publish and attach images. The agent prepares the content, drafts it, handles the metadata, and gives me a preview.
If something needs changing, I ask in the same conversation. Nothing goes live without my approval.
What interests me most is how this changes my relationship with the CMS.
Sanity still holds the content, but I no longer need to work through its interface for every update. I can describe what I want, review the result, and refine it.
As a Design Engineer, that gives me a different problem to think about: how to make the agent’s work easy to review, correct, and approve.
The preview and approval step matter as much as the ability to create a draft.
This is still a work in progress. I’m testing it on my own website and figuring out where a conversation makes the work easier—and where it needs more structure.
I’ll share more as it develops
An AI agent issues credits on disputed freight invoices. A concept that shows the cost of its mistakes before anyone clicks.
