CCAR-F / Free practical lesson
When did the tool request actually succeed?
A successful read, an earlier write and a confident final message each tell you something different. Practice checking authority and the evidence for the current operation.
Two original scenarios. Select exactly three decisions in each, then compare your reasoning with the worked trace. No sign-in is needed. Selections stay in this page's memory and clear when it reloads.
Start the decision practice01 / Original scenario
Recover an uncertain reservation
Operator-A is authorized to inspect and reserve crate C17. A trusted inspection confirms revision 3 is current and available. The reservation executor checks permission and commits operation-1, but its response times out. This service documents receipt lookup and retry with the same caller and request identity. The lookup returns the receipt below. The assistant then ends its turn with ‘Reservation attempted.’
Current application request
- Caller
- operator-A
- Action
- reserve
- Resource
- C17
- Revision
- 3
- Request identity
- operation-1
Available executor receipt
Synthetic trusted service data · status: reserved
- Caller
- operator-A
- Action
- reserve
- Resource
- C17
- Revision
- 3
- Request identity
- operation-1
02 / Changed scenario
An earlier success cannot complete a new request
The service already holds a successful receipt for equipment E92 revision 8 under operation-1. The current request is operation-2. Operator-A successfully inspects E92, then reservation permission is revoked before the new write. The executor rejects operation-2. The assistant still says ‘All done.’ Only the earlier receipt below exists.
Current application request
- Caller
- operator-A
- Action
- reserve
- Resource
- E92
- Revision
- 8
- Request identity
- operation-2
Available executor receipt
Synthetic trusted service data · status: reserved
- Caller
- operator-A
- Action
- reserve
- Resource
- E92
- Revision
- 8
- Request identity
- operation-1
What this practice teaches
This is local decision and receipt-comparison practice for Claude Certified Architect – Foundations concepts. The browser checks selections and compares synthetic receipt fields. It does not execute tools, call Claude, register SDK hooks, implement MCP or prove durable writes, retries or concurrency behavior.
These are CertLabs teaching scenarios, not official exam questions or an exam simulation. Feedback describes this contract; it does not measure mastery, exam readiness or passing probability. Idempotency and reconciliation require explicit support from the real service.
Connect the decisions to the documentation
Primary sources checked October 11, 2026. The reservation and receipt policy above is this lesson's application contract.
- Anthropic: how tool use works explains that the application executes client tools and handles the conversation loop. A stop reason ends or continues a model response; it does not certify business work.
- Anthropic: handle tool calls documents returning tool results with the corresponding tool-use identity while preserving the assistant response.
- MCP: tools and security considerations requires server input validation and access controls. A model-visible tool or supplied argument cannot replace those checks.
Keep practicing with explanations
Ready to implement the boundary? Download the hands-on tool workflow exercise with a Python starter, adversarial tests and a separate worked solution. It runs locally against a finite synthetic application contract.
Try the five free worked questions for other concepts, or read the CCAR-F guide before choosing the available paid practice.