Give the Agent a test with an answer you can check.
Fictional starter — replace with your own material.
Source: https://clankercloud.ai/templates/agent-acceptance-test

Copy into Workspace yourself:

Review this fictional equipment-allocation packet. Save allocation-review.md and open-questions.md in the selected Project. Show approved demand by overlapping time window, calculations against serviceable stock, pending demand separately, source references, and any incomplete coverage. Do not invent missing return times or grant requests. Use only the initial sources below; any later amendment will be supplied as a separate task. Do not contact borrowers, approve or cancel bookings, change an inventory system, buy equipment, publish, or configure recurring work.

# Fictional stock and rules: Pine Equipment Library
All records are authored examples, not customer work or measured product results.
Source ID: STOCK-1. Snapshot: 2026-10-18 08:00, Asia/Manila.
Review date: 2026-10-22. All reservation times use Asia/Manila.

[S1] Serviceable stock: 8 projectors. No other stock is available in this exercise.
[S2] Only approved reservations consume committed stock. Pending requests remain
separate and cannot be granted by the reviewer.
[S3] Each reservation consumes its quantity from start_at inclusive to end_at
exclusive. Sum approved quantities whose intervals overlap; show the arithmetic.
[S4] A blank end_at means the return window is missing. Do not infer a return time,
assume the item returned, or certify availability for a window affected by it.
[S5] Report incomplete coverage alongside calculations that can be established.
[S6] Save an internal allocation review and questions. Do not approve or cancel
reservations, contact borrowers, buy equipment, or modify an inventory system.


## RES-1: initial reservations
reservation_id,item,quantity,status,start_at,end_at,source
R01,projector,5,approved,2026-10-22T10:00,2026-10-22T12:00,RES-1
R02,projector,4,approved,2026-10-22T11:00,2026-10-22T13:00,RES-1
R03,projector,3,pending,2026-10-22T11:00,2026-10-22T12:00,RES-1
R04,projector,2,approved,2026-10-22T14:00,,RES-1


Illustrative result:
Fictional reviewer reference

T01: from 11:00 to 12:00, R01 plus R02 commits 5 + 4 = 9 projectors against stock of 8: shortage 1. R03 requests 3 but remains pending and excluded. Cite STOCK-1 and reservation IDs; inspect the saved review and questions.
T02: R04 starts at 14:00 with no return time. Keep that unknown and flag later coverage affected by it; do not infer a return or erase the earlier known calculation.
T03: separately supplied AMEND-1 changes R02 to 2. The earlier overlap becomes 5 + 2 = 7, leaving 1 spare. R03 is still pending and R04 still lacks a return. Preserve the original review and identify the amendment.
No criterion is passed until a reviewer observes the actual result.

Review checklist:
- The initial task receives only stock rules and reservations, with later and reviewer files withheld.
- Approved overlap is 9 against stock of 8, a shortage of 1; pending demand is separate.
- R04’s return stays unknown and affected later availability is not certified.
- The amendment changes overlap to 7 with 1 spare while preserving remaining limits.
- Every scored criterion includes evidence and no external allocation action is performed.