Prompt Agent
Prompt + input form + output template + knowledge references. No code execution. Ships fastest.
- Up in under an hour
- Open to all employees
- Knowledge Q&A / doc summary / email draft
Product · AiDo Studio
6-step wizard.
Let employees build their own agents.
The one-click agent builder. Frees colleagues from prompts, tools, configs and approvals — six steps to create, tune to taste, publish on demand and gate by approval.
Agent Types
Pick the form by scenario complexity. Skill floor rises stepwise.
Prompt covers 80% of cases, Workflow handles process automation, Code only when you really need code.
Workflow Agent
Multi-step workflow. Calls knowledge, OCR, approvals, notifications. Bound by tool whitelist.
- Visual step composition
- Human-in-the-loop nodes
- Approvals / inspections / automation
Code Agent
Arbitrary scripts and external APIs. Must run in sandbox. Resource quotas and audit always on.
- Python / Node sandbox
- Resource quota audit
- Data processing / complex integration
Wizard · 6 steps
Configure an agent like filling a form.
Each step has presets and examples. Worried about config? End the flow with a one-shot test.
Basics
Agent name, slug, description, icon, BU.
Persona
Target user, role positioning, tone, constraints.
Capabilities
Agent type, knowledge references, tool whitelist, model tier.
Sandbox & memory
Run sandbox, context window, long-term memory, sensitive-term policy.
Preview
One-shot test live, see output, adjust until satisfied.
Choose release scope
Private / team / company-wide, with corresponding approval flow.
Skill Hub · Enterprise Skill marketplace
A Skill shelf that belongs to your company.
Beyond the agent factory, Studio hosts an enterprise Skill marketplace. Employees turn their best capabilities into reusable Skills — the company gets a shelf where everyone can install, rate, version and iterate. A year of frontline practice becomes a one-click asset.
What is the Skill Hub? Skills are the minimum reusable unit of an Agent. Smaller and more focused than a full Agent — "doc summarizer", "contract clause extractor", "weather lookup", "inspection photo classifier", "OCR invoice to ledger" — these are typical Skills. After building an Agent in Studio, employees abstract its best parts into Skills and publish them to the enterprise Skill marketplace; other colleagues or departments install them with one click, no re-implementation needed.
The marketplace has ratings, install statistics, version control and submission review — paired with Studio's three-tier release model (private / team / company-wide). Low-quality or overreaching Skills are removed automatically; high-rated Skills are featured and join new-employee Agent templates. Each Skill carries an explicit permission declaration (tools, knowledge, data) so admins can control visibility and install rights by department and sensitivity tier — both rewarding employees for codifying experience and stating governance boundaries up front.
Global DT also publishes "industry Skill packs" (tourism, city, ecology, finance) as official Skills, so customers shift from "buying a platform once" to "continuously pulling new capability from the Skill Hub".
Cross-team reuse
Employees abstract the best single-purpose capabilities from their private agents into Skills. Finance, HR, ops, marketing — all install with one click. No wheels reinvented.
Skills are Lego for Agents
Skills are smaller than agents. 'Doc summarizer', 'contract clause extractor', 'weather lookup', 'inspection photo classifier' are all Skills. Agents compose Skills to finish jobs.
Rating / version / review
Built-in ratings, install counts, version control and review. Low-quality or overreaching Skills auto-removed. High-rated Skills land in recommended and new-hire agent templates.
Enterprise governance
Each Skill carries a permission declaration (tools, knowledge, data). Admins control visibility and install rights per department and sensitivity tier.
Release Tiers
Different scope, different approval.
An Agent doesn't need to be 'visible to all immediately'. Run private first, then team, then company-wide — every tier has its own flow.
Private release
Only you can use it. No approval. Create, iterate, refine on your own first.
Team release
After team lead approval, visible inside the team. Lands in team recommendations.
Company-wide
Enters company Agent Store. Platform admin reviews + high-risk tool re-authorization.
Capabilities
Studio isn't an island — it sits on AI-OS.
Knowledge, tools, models, quotas, audit are not built by Studio. They come straight from AI-OS.
Knowledge space
Agents mount enterprise knowledge bases, vector indexes and graphs directly. Q&A out of the box.
Tool whitelist
Email, docs, SaaS, private APIs, internal tickets — tool set scoped by agent type.
Model tier choice
Instant / Pro across unified model pool, switch by complexity.
Run quota audit
Call count, concurrency, token usage bound by tenant. Every call audited.
FAQ
Studio vs AiDo / safety / IT requirements / models / direct try — five common questions.
Studio vs AiDo — what's the difference?
AiDo Studio is the agent factory employees use to build agents. AiDo is the agent suite — ready-made agents (task assistant, doc assistant, on-call assistant). Studio creations one-click publish into AiDo Store.
Are employee-built agents safe?
Three lines of defense: ① tool whitelist controls what they can do; ② three-tier release with approvals; ③ quota, audit, sensitive-term filters always on. Platform admin can take down any agent. Code agents must sandbox.
What IT does Studio require?
Depends on AI-OS identity, knowledge space and model gateway. Deploys on public, dedicated or private cloud. Code agent sandbox is recommended inside the corporate network, integrating with existing SSO and monitoring.
Which models? Can we plug in our own?
Models flow through the AI-OS gateway. Instant / Pro by default. Customers can mount private models (vLLM / TensorRT / Qwen / Ernie etc.) per agent.
Try it directly?
Yes. Visit aios.shukeyun.com/aido-studio, sign in with your enterprise email, create your first Prompt Agent. We recommend starting with an internal Q&A bot before moving to Workflow / Code agents.
Start now
Ship your first Prompt Agent first.
Visit the AiDo workshop, build your first Prompt Agent in 15-30 minutes. Need more complex Workflow / Code agents? Get in touch directly.