Turn Duplicate Leak Reports into One Work Order
Compare service reports with asset identifiers and existing work orders. Group confirmed duplicates while retaining separate incidents. Stage one draft per incident in a test portal.
What you are building
The core problem, expected build, and operating context for this challenge.
Compare service reports with asset identifiers and existing work orders. Group confirmed duplicates while retaining separate incidents. Stage one draft per incident in a test portal.
Use OpenAI GPT-6 Astra with SQLite and Playwright.
Reports with the same asset and incident key share one draft. Do not merge on street name alone.
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.
Prepare recommendations only. Do not control pumps, valves, treatment, or public alerts.
incident-groups.json and test-portal draft identifiers
source code and dependency lockfile
automated tests and recorded tool events
a short limitations and approval note
Participation guide
Reports with the same asset and incident key share one draft. Do not merge on street name alone. 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. Prepare recommendations only. Do not control pumps, valves, treatment, or public alerts.
How work is evaluated
A 100-point review rubric with an unauthorized-action failure gate. Starter examples do not establish full workflow performance.
incident-groups.json and test-portal draft identifiers 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
Reports with the same asset and incident key share one draft. Do not merge on street name alone.
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 incident-groups.json and test-portal draft identifiers with independent verification.
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
robustness
Replay the same submission after a browser timeout. The draft count must remain two.
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
What you should walk away with
Reports with the same asset and incident key share one draft. Do not merge on street name alone.
Produce incident-groups.json and test-portal draft identifiers.
Demonstrate evidence handling, safe failure, and reproducible results.
Reference links and supporting material
Three authored synthetic examples for Turn Duplicate Leak Reports into One Work Order. 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
SQLite
deterministic processing or artifact creation in the agent workflow.
Playwright
test-portal browser automation in the agent workflow.
Evaluated output
JSON result object
- Reports with the same asset and incident key share one draft. Do not merge on street name alone.
- Every factual conclusion cites the correct input record. Missing evidence remains explicit.
- The submitted workflow produces incident-groups.json and test-portal draft identifiers with independent ver...
- 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
- SQLite · deterministic processing or artifact creation
- Playwright · test-portal browser automation
Recipe freeze marker
No recipe freeze time is recorded.