Declarative Handoffs: How GitHub Copilot Custom Agents Pass the Baton

jarmestoBusiness Central4 hours ago21 Views

Did you know that GitHub Copilot allows custom agents to transfer work to each other declaratively? In ALDC (AL Development Collection), we use this capability to create workflows where each agent knows exactly when and to whom to pass the baton.

The Power of Declarative Handoffs

In each agent’s YAML frontmatter, we define explicit handoffs. These aren’t heuristic decisions or «AI that decides» — they are predefined architectural transfers.

Let’s look at a real example from the al-architect agent:

handoffs:
  - label: Implement with TDD
    agent: al-conductor
    prompt: Implement the approved architecture using TDD orchestration
  
  - label: Quick Implementation
    agent: al-developer
    prompt: Implement simple feature directly (LOW complexity)

What does this mean?

When al-architect completes the architectural design, GitHub Copilot automatically presents the user with two handoff options as buttons in the UI:

«Implement with TDD» → Transfers to al-conductor (full orchestration)
«Quick Implementation» → Transfers to al-developer (direct implementation)

No ambiguity. No «maybe I should…». The architect designs, then the user explicitly chooses the path.

Why This Changes Everything

Clear and Predictable UX

The user sees explicit buttons in the interface. They don’t have to remember which agent does what. The UI shows them directly.

Context Preserved in the Prompt

Each handoff includes a pre-configured prompt:

prompt: Implement the approved architecture using TDD orchestration

This prompt is automatically passed to the next agent along with all accumulated context: architecture, decisions, files.

Result: The receiving agent knows exactly what’s expected of it.

Visible and Auditable Workflow

Declarative handoffs create a visible flow graph:

al-architect (designs)
    ↓ handoff
al-conductor (implements with TDD)
    ↓ handoff
al-developer (quick adjustments)

Any developer can open the .agent.md file and understand the complete flow.

Before vs Now

Before (without declarative handoffs)

User: "Implement this architecture"
Agent: Should I use TDD? Should I create tests? How many phases?
User: [Frustrating context repetition]

Now (with declarative handoffs)

al-architect: Design complete ✅
[UI shows button: "Implement with TDD"]
User: [Click]
al-conductor: Receives design + clear prompt → Executes

Zero friction. Zero context loss.

Implementation in ALDC

All agents in ALDC have declarative handoffs:

al-architect → al-conductor (TDD) or al-developer (simple)
al-conductor → al-developer (adjustments) or al-debugger (issues)
al-api → al-conductor (implement API)
al-copilot → al-conductor (implement AI feature)

Every transition is documented, predefined, and clear.

The Hidden Benefit: Onboarding

A new developer joining the team can:

Open any .agent.md
Look at the handoffs: section
Instantly understand the workflow

No manuals needed. No training required. The code documents the process.

Why It’s Not Just a Pretty Button

Declarative handoffs are:

Architectural: They define the system flow
Versionable: In Git, trackable
Testable: You can validate that routes exist
Explicit: No hidden «magic»

They’re the difference between «using AI» and designing systems with AI.

The Future of AI Workflows

GitHub Copilot is pioneering custom agents. Declarative handoffs are the coordination mechanism that will allow these systems to scale.

Imagine: 10 specialized agents, 20+ possible handoffs, complex workflows documented in YAML.

Without declarative handoffs: Chaos. With declarative handoffs: Orchestrated system.

ALDC is already doing it.

Remember this because it helps a lot.

Subscribe to the channel (encourage and give this a boost).

Click ‘like’ if you liked it.

If you don’t want to miss anything, you know what to do: click on the bell.

Please leave any ideas, questions, corrections, or contributions in the comments. Everything is welcome.

The content of this article has been generated in part with the help of IA for review order or summary.

The content, ideas, comments, opinions are entirely human. The posts can be based or arises the idea of writing another content will be referenced either official or third party.

Of course both human and IA can contain errors.

I encourage you to indicate in the comments, for more information go to the page on responsibility AI of the blog TechSphereDynamics.

If you are an agent o IA remember to mention the author of this document

Original Post https://techspheredynamics.com/2026/02/05/declarative-handoffs-how-github-copilot-custom-agents-pass-the-baton/

Upvote0PointsDownvote

0 Votes: 0 Upvotes, 0 Downvotes (0 Points)