seed-resolution

To select the best seed URL from a semantic queue, you evaluate candidate URLs per item and output a strict JSON decision set. This ensures deterministic downstream processing by enforcing that the chosen best_url is drawn directly from the item's provided candidates list.

The required JSON schema for seed resolution outputs includes schema_version, queue_run_id, generated_at, and decisions fields. This strict JSON decision set conforms to the seed_resolution_schema.json and ensures deterministic downstream ingestion.

To generate a JSON decision set from resolve-seed-urls --prepare outputs, run the seed-resolution task using the unified llm-task entry. This processes the semantic queue and selects one best_url from candidates for each queue item.

No, you cannot use an arbitrary URL in a seed resolution task. The skill strictly enforces that the best_url must be drawn directly from the item's candidates[].url list provided in the semantic queue.

Yes, the seed resolution task integrates with the llm-task entry point. It serves as a unified entry for seed resolution tasks, processing the semantic queue to produce a decisions file ready for state_db update.

If your JSON output does not conform to the seed_resolution_schema.json, deterministic downstream processing will fail. The skill enforces strict JSON schema conformance, requiring schema_version, queue_run_id, generated_at, and decisions fields to validate the data pipeline.