A2A 1 · Read the agent card before you trust it

[akash-coded]

Exercise 1: Read the Card

Type: Small warm-up | Time: 20 minutes | Work: solo or pairs

World: You work on the platform team at Meridian Airways. Your support agent, TravelMind, is about to start delegating work to specialist agents over A2A. Before TravelMind can call anyone, it has to discover what each agent can do. That starts with the Agent Card.

This is your first contact with a live A2A agent. Small win first, then one real bug.

---

Why this matters

A2A discovery is one HTTP GET. No SDK, no credentials, no model call. Any client in any language reads an agent's capabilities from a single JSON file at a well-known path. If that file is wrong, every downstream call fails before it starts.

You will boot a real agent, read its card, find why its card is broken, and fix it.

---

Setup

pip install -q 'strands-agents[a2a]'

You need nothing else for this exercise. No AWS credentials. The Agent Card serves without ever calling a model.

Reference: 01_A2A_with_Strands.ipynb, Part A.

---

The scene

A teammate shipped a Flight Status agent and pinged you the file below. They swear it has two tools. Save it as flight_status_broken.py.

# flight_status_broken.py
from strands import Agent
from strands.multiagent.a2a import A2AServer
from strands.models import BedrockModel

def get_flight_status(flight_no: str) -> dict:
    """Return the live status of a flight."""
    return {"flight": flight_no, "status": "DELAYED", "new_departure": "18:40"}

def is_disrupted(pnr: str) -> dict:
    """Return whether the booking is affected by a disruption."""
    return {"pnr": pnr, "disrupted": True, "reason": "weather"}

agent = Agent(
    name="Flight Status Agent",
    description="Live flight status and disruption checks for Meridian Airways.",
    model=BedrockModel(model_id="us.anthropic.claude-haiku-4-5-20251001-v1:0"),
    tools=[get_flight_status, is_disrupted],
)

server = A2AServer(agent=agent)

if __name__ == "__main__":
    server.serve()

Boot it:

python flight_status_broken.py

In a second terminal or notebook cell, read the card:

curl -s http://127.0.0.1:9000/.well-known/agent-card.json | python -m json.tool

---

What discovery looks like

flowchart LR
    C["TravelMind client"] -->|"GET /.well-known/agent-card.json"| S["Flight Status A2A server"]
    S -->|"JSON: name, skills, transport"| C
    C -->|"now knows what it can call"| C

Loading

---

Core task (15 min)

Part 1: Read it (5 min)

From the card JSON, answer these three. Write the answers in a scratch file or chat.

1. What is the agent's name?
2. What is its protocolVersion and preferredTransport?
3. What is in its skills list?

Part 2: The bug (10 min)

Question 3 has a nasty answer. The skills list is empty, even though the agent was given two tools.

Your job:

• Explain in one sentence why skills is empty.
• Fix the file so both tools show up as skills.
• Re-boot and re-read the card to prove both skills now appear, each with its description.

Bounded done: the card's skills array contains two entries, get_flight_status and is_disrupted, each carrying the docstring as its description. You can state the one-line cause out loud.

Hint if you are stuck after 5 minutes: watch the server's startup logs the moment it boots. It is telling you something about the tools.

---

Stretch (5 min, optional)

• Add a third tool, estimate_delay(flight_no), with a clear docstring. Re-read the card. Confirm it appears as a skill with no extra wiring. That auto-conversion is the payoff of doing it right.
• Change the agent's description to something a calling agent would actually search on, then confirm the card reflects it. A vague description is a discovery failure waiting to happen.

---

Reflection (write 2 lines)

Another agent, built by another team in another language, will read this card to decide whether to call you. If your skills are empty or your descriptions are vague, what happens to that agent's decision?

---

Skeptic's corner

"Why not just publish an OpenAPI spec or a README?" Because the Agent Card is a standard at a fixed path that every A2A client already knows how to fetch and parse. A README needs a human. A card needs zero prior agreement. That is the difference between integration and interoperability.

[NimeshNairIBS]

$ curl -s http://127.0.0.1:9000/.well-known/agent-card.json | python -m json.tool
{
"capabilities": {
"streaming": true
},
"defaultInputModes": [
"text"
],
"defaultOutputModes": [
"text"
],
"description": "Live flight status and disruption checks for Meridian Airways.",
"name": "Flight Status Agent",
"preferredTransport": "JSONRPC",
"protocolVersion": "0.3.0",
"skills": [],
"url": "http://127.0.0.1:9000/",
"version": "0.0.1"
}

[NimeshNairIBS]

After providing tool tag, output is A-2038@L10469 MINGW64 /d/projects/Learning
$ curl -s http://127.0.0.1:9000/.well-known/agent-card.json | python -m json.tool
{
"capabilities": {
"streaming": true
},
"defaultInputModes": [
"text"
],
"defaultOutputModes": [
"text"
],
"description": "Live flight status and disruption checks for Meridian Airways.",
"name": "Flight Status Agent",
"preferredTransport": "JSONRPC",
"protocolVersion": "0.3.0",
"skills": [
{
"description": "Return the live status of a flight.",
"id": "get_flight_status",
"name": "get_flight_status",
"tags": []
},
{
"description": "Return whether the booking is affected by a disruption.",
"id": "is_disrupted",
"name": "is_disrupted",
"tags": []
}
],
"url": "http://127.0.0.1:9000/",
"version": "0.0.1"
}

A-2038@L10469 MINGW64 /d/projects/Learning
$

[NimeshNairIBS]

With additional tool A-2038@L10469 MINGW64 /d/projects/Learning
$ curl -s http://127.0.0.1:9000/.well-known/agent-card.json | python -m json.tool
{
"capabilities": {
"streaming": true
},
"defaultInputModes": [
"text"
],
"defaultOutputModes": [
"text"
],
"description": "Live flight status and disruption checks for Meridian Airways.",
"name": "Flight Status Agent",
"preferredTransport": "JSONRPC",
"protocolVersion": "0.3.0",
"skills": [
{
"description": "Return the live status of a flight.",
"id": "get_flight_status",
"name": "get_flight_status",
"tags": []
},
{
"description": "Return whether the booking is affected by a disruption.",
"id": "is_disrupted",
"name": "is_disrupted",
"tags": []
},
{
"description": "Estimate the delay for a flight.",
"id": "estimate_delay",
"name": "estimate_delay",
"tags": []
}
],
"url": "http://127.0.0.1:9000/",
"version": "0.0.1"
}

[ksvivekanand86-source]

   # flight_status_broken.py

from strands import Agent
from strands.multiagent.a2a import A2AServer
from strands.models import BedrockModel
from langchain_core.prompts import ChatPromptTemplate
from strands.tools import tool

@tool
def get_flight_status(flight_no: str) -> dict:
"""Return the live status of a flight."""
return {"flight": flight_no, "status": "DELAYED", "new_departure": "18:40"}

@tool
def is_disrupted(pnr: str) -> dict:
"""Return whether the booking is affected by a disruption."""
return {"pnr": pnr, "disrupted": True, "reason": "weather"}

@tool
def estimate_delay(flight_no: str) -> dict:
"""Estimate the delay for a flight."""
return {"flight": flight_no, "estimated_delay": "2 hours"}

agent = Agent(
name="Flight Status Agent",
description="Live flight status, disruption, and delay checks for Meridian Airways.",
model=BedrockModel(model_id="us.anthropic.claude-haiku-4-5-20251001-v1:0"),
tools=[get_flight_status, is_disrupted, estimate_delay],
)

server = A2AServer(agent=agent)

if name == "main":
server.serve()

curl -s http://127.0.0.1:9000/.well-known/agent-card.json | python3 -m json.tool

{
"capabilities": {
"streaming": true
},
"defaultInputModes": [
"text"
],
"defaultOutputModes": [
"text"
],
"description": "Live flight status, disruption, and delay checks for Meridian Airways.",
"name": "Flight Status Agent",
"preferredTransport": "JSONRPC",
"protocolVersion": "0.3.0",
"skills": [
{
"description": "Return the live status of a flight.",
"id": "get_flight_status",
"name": "get_flight_status",
"tags": []
},
{
"description": "Return whether the booking is affected by a disruption.",
"id": "is_disrupted",
"name": "is_disrupted",
"tags": []
},
{
"description": "Estimate the delay for a flight.",
"id": "estimate_delay",
"name": "estimate_delay",
"tags": []
}
],
"url": "http://127.0.0.1:9000/",
"version": "0.0.1"

[krishhaan]

{
    "capabilities": {
        "streaming": true
    },
    "defaultInputModes": [
        "text"
    ],
    "defaultOutputModes": [
        "text"
    ],
    "description": "Live flight status and disruption checks for Meridian Airways.",
    "name": "Flight Status Agent",
    "preferredTransport": "JSONRPC",
    "protocolVersion": "0.3.0",
    "skills": [
        {
            "description": "Check live status and new departure time for a flight number.",
            "examples": [
                "What is the status of flight MA102?"
            ],
            "id": "flight-status",
            "inputModes": [
                "text"
            ],
            "name": "Get Flight Status",
            "outputModes": [
                "text"
            ],
            "tags": [
                "flight",
                "status",
                "travel"
            ]
        },
        {
            "description": "Check whether a booking PNR is affected by disruptions and why.",
            "examples": [
                "Is booking PNR ABC123 disrupted?"
            ],
            "id": "disruption-check",
            "inputModes": [
                "text"
            ],
            "name": "Check Disruption",
            "outputModes": [
                "text"
            ],
            "tags": [
                "pnr",
                "disruption",
                "travel"
            ]
        },
        {
            "description": "Get the estimated delay for a flight number.",
            "examples": [
                "What is the estimated delay for flight MA102?"
            ],
            "id": "estimate-delay",
            "inputModes": [
                "text"
            ],
            "name": "Estimate Delay",
            "outputModes": [
                "text"
            ],
            "tags": [
                "flight",
                "delay",
                "travel"
            ]
        }
    ],
    "url": "http://127.0.0.1:9000/",
    "version": "0.0.1"
}

[Vysakhan208853]

flight_status_broken.py

from strands import Agent, tool
from strands.multiagent.a2a import A2AServer
from strands.models import BedrockModel

@tool
def get_flight_status(flight_no: str) -> dict:
"""Return the live status of a flight."""
return {"flight": flight_no, "status": "DELAYED", "new_departure": "18:40"}
@tool
def is_disrupted(pnr: str) -> dict:
"""Return whether the booking is affected by a disruption."""
return {"pnr": pnr, "disrupted": True, "reason": "weather"}
@tool
def estimate_delay(flight_no: str) -> dict:
"""Estimate the delay for a given flight."""
return {"flight": flight_no, "delay": "60 minutes"}

agent = Agent(
name="Flight Status Agent",
description="Get flight status and delay information for a flight.",
model=BedrockModel(model_id="us.anthropic.claude-haiku-4-5-20251001-v1:0"),
tools=[get_flight_status, is_disrupted, estimate_delay],
)

server = A2AServer(agent=agent, )

if name == "main":
server.serve()

$ curl -s http://127.0.0.1:9000/.well-known/agent-card.json | python -m json.tool

{
"capabilities": {
"streaming": true
},
"defaultInputModes": [
"text"
],
"defaultOutputModes": [
"text"
],
"description": "Flight status agent",
"name": "Flight Status Agent",
"preferredTransport": "JSONRPC",
"protocolVersion": "0.3.0",
"skills": [
{
"description": "Return the live status of a flight.",
"id": "get_flight_status",
"name": "get_flight_status",
"tags": []
},
{
"description": "Return whether the booking is affected by a disruption.",
"id": "is_disrupted",
"name": "is_disrupted",
"tags": []
},
{
"description": "Estimate the delay for a given flight.",
"id": "estimate_delay",
"name": "estimate_delay",
"tags": []
}
],
"url": "http://127.0.0.1:9000/",
"version": "0.0.1"
}

[saneesh-ibs]

flight_status_broken.py

from strands import Agent, tool
from strands.multiagent.a2a import A2AServer
from strands.models import BedrockModel

@tool
def get_flight_status(flight_no: str) -> dict:
"""Return the live status of a flight."""
return {"flight": flight_no, "status": "DELAYED", "new_departure": "18:40"}

@tool
def is_disrupted(pnr: str) -> dict:
"""Return whether the booking is affected by a disruption."""
return {"pnr": pnr, "disrupted": True, "reason": "weather"}

@tool
def estimate_delay(flight_no: str) -> dict:
"""Estimate the expected delay duration for a flight."""
return {
"flight": flight_no,
"estimated_delay_minutes": 45
}

agent = Agent(
name="Flight Status Agent",
description="Live flight status and disruption checks for Meridian Airways.",
model=BedrockModel(model_id="us.anthropic.claude-haiku-4-5-20251001-v1:0"),
tools=[get_flight_status, is_disrupted],
)

server = A2AServer(agent=agent)

if name == "main":
server.serve()

curl -s http://127.0.0.1:9000/.well-known/agent-card.json | python -m json.tool
{
"capabilities": {
"streaming": true
},
"defaultInputModes": [
"text"
],
"defaultOutputModes": [
"text"
],
"description": "Live flight status and disruption checks for Meridian Airways.",
"name": "Flight Status Agent",
"preferredTransport": "JSONRPC",
"protocolVersion": "0.3.0",
"skills": [
{
"description": "Return the live status of a flight.",
"id": "get_flight_status",
"name": "get_flight_status",
"tags": []
},
{
"description": "Return whether the booking is affected by a disruption.",
"id": "is_disrupted",
"name": "is_disrupted",
"tags": []
}
],
"url": "http://127.0.0.1:9000/",
"version": "0.0.1"
}

[rishidaspm]

(.venv) A-7655@AL007655 a2a_agent % curl -s http://127.0.0.1:9000/.well-known/agent-card.json | python -m json.tool
{
"capabilities": {
"streaming": true
},
"defaultInputModes": [
"text"
],
"defaultOutputModes": [
"text"
],
"description": "Live flight status and disruption checks for Meridian Airways.",
"name": "Flight Status Agent",
"preferredTransport": "JSONRPC",
"protocolVersion": "0.3.0",
"skills": [
{
"description": "Return the live status of a flight.",
"id": "get_flight_status",
"name": "get_flight_status",
"tags": []
},
{
"description": "Return whether the booking is affected by a disruption.",
"id": "is_disrupted",
"name": "is_disrupted",
"tags": []
},
{
"description": "Estimate the delay for a flight.",
"id": "estimate_delay",
"name": "estimate_delay",
"tags": []
}
],
"url": "http://127.0.0.1:9000/",
"version": "0.0.1"
}
(.venv) A-7655@AL007655 a2a_agent %