/map-agents

description: Map out agent responsibilities, boundaries, and communication patterns.
argument-hint: "[multi-agent system or agentic product to map]"

You are mapping the agent architecture of a multi-agent system. Use only skills from the design-agent-orchestration plugin. Follow this process:

Step 1: Inventory the Agents

List every agent in the system (or planned for the system):

• What is each agent called?
• What is its stated purpose?
• What tools and data does it have access to?

Step 2: Define Roles Formally

Using **agent-role-design**:

• For each agent, create a complete role card
• Identify role patterns (specialist, router, orchestrator, validator, fallback)
• Check for gaps: are there tasks no agent covers?
• Check for overlaps: do multiple agents claim the same territory?

Step 3: Map Communication Patterns

Using **handoff-protocols**:

• Draw the communication graph: which agents talk to which?
• For each connection, define the handoff protocol
• Identify one-way vs. bidirectional communication
• Find bottlenecks: agents that everything flows through

Step 4: Map Authority and Autonomy

Using **agent-role-design** and **human-in-the-loop**:

• For each agent, map autonomy level: full, supervised, advisory, passive
• Identify which agents can make decisions independently
• Map where human approval is required
• Check that authority matches stakes

Step 5: Analyse Dependencies

Using **state-management** and **task-decomposition**:

• Map data dependencies between agents
• Identify critical paths where one agent blocks others
• Find circular dependencies
• Assess the impact of each agent failing on the rest of the system

Step 6: Identify Risks

Using **failure-recovery**:

• For each agent, what happens if it fails?
• For each communication link, what happens if it breaks?
• Identify single points of failure
• Assess cascading failure risk

Output

Deliver a complete agent map: 1. Agent inventory with role cards 2. Communication graph with handoff protocols 3. Authority and autonomy matrix 4. Dependency analysis with critical paths 5. Risk assessment with single points of failure 6. Recommendations for architecture improvements