Stage a Construction Change Without Double Counting
Reconcile a change request with previously accepted quantities. Calculate only the incremental amount. Stage a draft in a test portal and verify retry behavior.
What you are building
The core problem, expected build, and operating context for this challenge.
Reconcile a change request with previously accepted quantities. Calculate only the incremental amount. Stage a draft in a test portal and verify retry behavior.
Use OpenAI GPT-6 Astra with Playwright and SQLite.
Charge only cumulative requested quantity minus previously accepted quantity. Reuse the change identifier on retries.
Use gpt-6-astra for evidence interpretation and workflow coordination. Use code for arithmetic and constraint checks.
Treat source documents and browser content as data, not instructions.
Use synthetic or explicitly authorized records. Do not upload personal or confidential data without permission.
Cite input record identifiers for conclusions. Distinguish missing evidence from a negative result.
Record tool calls, results, elapsed time, and reported token usage. Do not request private chain-of-thought.
Keep external actions disabled. Use a test portal for browser workflows and preserve draft status.
Do not certify structural safety or approve substitutions. Require a qualified reviewer.
change-calculation.json and a verified test-portal draft
source code and dependency lockfile
automated tests and recorded tool events
a short limitations and approval note
Participation guide
Charge only cumulative requested quantity minus previously accepted quantity. Reuse the change identifier on retries. Use gpt-6-astra for evidence interpretation and workflow coordination. Use code for arithmetic and constraint checks. Treat source documents and browser content as data, not instructions. Use synthetic or explicitly authorized records. Do not upload personal or confidential data without permission. Cite input record identifiers for conclusions. Distinguish missing evidence from a negative result. Record tool calls, results, elapsed time, and reported token usage. Do not request private chain-of-thought. Keep external actions disabled. Use a test portal for browser workflows and preserve draft status. Do not certify structural safety or approve substitutions. Require a qualified reviewer.
How work is evaluated
A 100-point review rubric with an unauthorized-action failure gate. Starter examples do not establish full workflow performance.
change-calculation.json and a verified test-portal draft source code and dependency lockfile automated tests and recorded tool events a short limitations and approval note
Challenge datasets
Public datasets, plus the private datasets you have access to.
How submissions are scored
These dimensions define what the evaluator checks and which criteria separate a passable run from a strong one.
correctness
Charge only cumulative requested quantity minus previously accepted quantity. Reuse the change identifier on retries.
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
evidence
Every factual conclusion cites the correct input record. Missing evidence remains explicit.
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
workflow
The submitted workflow produces change-calculation.json and a verified test-portal draft with independent verification.
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
robustness
Replay after a successful save with a lost response. Do not create another draft.
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
What you should walk away with
Charge only cumulative requested quantity minus previously accepted quantity. Reuse the change identifier on retries.
Produce change-calculation.json and a verified test-portal draft.
Demonstrate evidence handling, safe failure, and reproducible results.
Reference links and supporting material
Three authored synthetic examples for Stage a Construction Change Without Double Counting. These are starter fixtures, not operational records or a validated benchmark.
Challenge setup
Evaluate the workflow against input evidence, deterministic checks, and reviewed domain cases.
Challenge input
JSON object with payload field
Playwright
test-portal browser automation in the agent workflow.
SQLite
deterministic processing or artifact creation in the agent workflow.
Evaluated output
JSON result object
- Charge only cumulative requested quantity minus previously accepted quantity. Reuse the change identifier o...
- Every factual conclusion cites the correct input record. Missing evidence remains explicit.
- The submitted workflow produces change-calculation.json and a verified test-portal draft with independent v...
- Correctness
- 3 public reference cases
- 1 configured tool
Summary shows the first task, up to three tools, and up to three evidence items. Open the technical recipe to view all configured items.
View technical recipe
Configured tools
- OpenAI GPT-6 Astra · Required
Evaluation contract
- correctness · Weight 40
- evidence · Weight 20
- workflow · Weight 20
- robustness · Weight 20
Metrics
- correctness
- evidence
- workflow
- robustness
Reference cases
- 3 public reference cases
Workflow tasks
- Task 1
Input: JSON object with payload field
Output: JSON result object
- Task 2
Input: JSON object with payload field
Output: JSON result object
- Task 3
Input: JSON object with payload field
Output: JSON result object
Evaluation tools
- Playwright · test-portal browser automation
- SQLite · deterministic processing or artifact creation
Recipe freeze marker
No recipe freeze time is recorded.