Stage 4 of 4. Read the project prerequisites before starting; this stage builds on the preceding contracts.
What changes
A bounded agent needs more than a money counter. Apply a deadline and a cost limit before each sequential job, then append an explicit completed or rejected event. Test deterministic durations rather than sleeping.
This is a simulator for admission decisions, clearly separated from execution. The JSON trace lets you inspect which constraint rejected each job. It is not a benchmark of model latency.
Work through one concrete case
Two jobs each reserve 70 against limit 100. In execution mode, the first settles 20; the second now fits and settles 20. In conservative replay with costs 70 and 70, only the first fits. The different traces follow different evidence.
pj agent budget planner 4
Change the lab inputs and calculate the result before reading its metrics. The figure computes from those inputs; the implementation tests below remain the source of completion evidence.
Implement the contract
Implement schedule against the stated contract. Emit each outcome in events[].status: completed for an admitted job, or rejected with its deadline or budget reason. The execution driver uses the same status field and adds needs_reconciliation for uncertain receipts.
Use the public API contract and the typed starter signatures. Return values from core functions and let the supplied driver own file input, argument parsing and presentation.
Keep schedule as a deterministic admission replay. The supplied execute_jobs driver calls your reserve and settle functions around an actual callback; do not substitute predicted duration for the callback's elapsed clock.
Verify and inspect
From the repository root, initialize once with python3 scripts/project_test.py agent-budget-planner --init learning-artifacts/agent-budget-planner. Then grade cumulatively:
python3 scripts/project_test.py agent-budget-planner --stage 4 --path learning-artifacts/agent-budget-planner --strictA fresh workspace should fail until you implement the contract. After every stage is complete, run your actual artifact from the supplied sample:
cd learning-artifacts/agent-budget-planner
python3 cli.py samples/input.json --mode execute --output budget.jsonInvestigate the failure boundary
Add an invocation that sleeps past the deadline. The next dispatch must stop, but this synchronous implementation cannot interrupt the already-running callback.
Costs are caller-supplied integer receipts. A monotonic deadline gates dispatch but cannot interrupt a synchronous callback. The ledger is single-process and in-memory.