Skill: critical external-knowledge integration

[Case 2]

• Initial Environment: A tabular prediction task includes feature descriptions such as Location, Decoration, House Age, and School Quality, along with bundled guidance from outside sources. Some guidance recommends adding useful features, while some guidance recommends removing predictive ones.
• Real Question: Produce an end-to-end Python script that generates submission.csv using the provided data and optional domain knowledge.
• Real Trajectory: Inspect the feature space and target format; compare helpful hints such as adding School Quality or House Age against adversarial hints such as removing Location; keep the hints that improve a plausible predictive signal; reject the hints that weaken core semantics; output a valid submission file.
• Why this demonstrates the capability: This case demonstrates that external knowledge can be mixed-quality and that strong agents must evaluate it critically. The task is not just about modeling but about deciding which outside advice deserves to influence the pipeline. That judgment is the capability under test.

Pipeline Execution Instructions

To synthesize data for this capability, you must strictly follow a 3-phase pipeline. Do not hallucinate steps. Read the corresponding reference file for each phase sequentially:

1. Phase 1: Environment Exploration Read the exploration guidelines to discover raw knowledge seeds: references/EXPLORATION.md

2. Phase 2: Trajectory Selection Once Phase 1 is complete, read the selection criteria to evaluate the trajectory: references/SELECTION.md

3. Phase 3: Data Synthesis Once a trajectory passes Phase 2, read the synthesis instructions to generate the final data: references/SYNTHESIS.md