Subagents

Subagents help RIB investigate complex tasks faster by splitting research, code exploration, asset generation, and review into focused work.

When a task needs more investigation, RIB can start temporary subagents to research, inspect, or review focused parts of the work. Each subagent works independently and, when useful, RIB can run several in parallel.

Subagents report their findings back to the main RIB agent. The main agent uses those findings to decide what to do next, continue building, or explain the result.

You do not need to configure subagents or choose when they run. RIB decides when subagents are useful based on your request.

Why subagents help

Subagents help the main RIB agent by giving it focused evidence before it acts.

They are useful when a request has several independent areas to inspect:

• One subagent can map relevant files.
• Another can review risk or architecture.
• Another can investigate a bug path.
• The main agent can compare the findings and choose a safer next step.

This keeps larger work from becoming one broad, unfocused search.

Current subagent types

Current subagent capabilities include:

Type	What it does	Can it change your project?
Investigation subagent	Searches, reads, maps relevant files, reviews architecture, and reports risks.	No. It is read-only.
Image generation subagent	Generates a project image asset for visual work when RIB needs one.	It can create an image asset, but the main RIB agent decides how to use it.

Investigation subagents cannot edit files, run commands, build previews, publish, restore versions, or make the final decision for the task.

Image generation is a special case for visual assets. Treat it as asset creation, not a general code-editing subagent.

What you may see in the chat

When the UI exposes subagent activity, you may see tool or activity cards showing that RIB is investigating something in the background.

Those cards are there to make the work visible. You do not need to manage them manually.

When subagents are useful

Subagents are useful when one request contains independent questions, such as:

• Review the current project structure.
• Inspect a payment flow.
• Check a visual layout issue.
• Compare implementation options.
• Investigate a bug without editing files.
• Generate a visual asset for a page or product experience.

The goal is focus. Each subtask should have a clear question and a clear output.

They help when work can be split safely:

• Research one area while another area is reviewed.
• Compare multiple files or flows.
• Investigate independent risks before building.
• Produce concise findings for the main agent to use.

When not to use subagents

Avoid splitting work when:

• The task requires one sequential edit.
• The project state is changing rapidly.
• You need a simple copy or layout change.
• The subtask depends on another unfinished subtask.

Copyable prompts

Use subagents to investigate why my dashboard is sometimes slow.
Look through the dashboard and anything connected to it, then tell me what might be causing the slowdown and how to improve it.
Do not edit files until you have reported the findings.

Before building, investigate this with focused subagents.
Check the current pages, backend/data dependencies, auth requirements, and risks.
Then summarize what should be changed first.

How to review subagent findings

Look for:

• Evidence from the project.
• Clear risks.
• No unsupported assumptions.
• A short recommendation.
• A next step that can be executed safely.

FAQ

Do I need to prompt differently?

No. You can ask normally. If you want deeper investigation, say that you want RIB to investigate first before editing.

Can subagents edit my code?

Investigation subagents cannot edit your code. They can search, inspect, research, and review, but they cannot create, edit, or delete files.

The main RIB agent is still responsible for project code changes.

Can multiple subagents run at once?

Yes. When a request has several independent parts, RIB can run multiple subagents in parallel.

Will subagents make my builds more expensive?

Subagents consume model usage like other agent work. RIB routes routine research and lookups to appropriate models where possible, which helps keep usage efficient.

Are subagents available on all plans?

Subagents are part of how RIB works and do not require setup from you. Availability can still depend on the product surface, plan, and request type.

Next steps

• Use planning for larger work: Plan mode
• Execute focused changes: Build mode (Agent)
• Manage reusable guidance: Skills