select-model

Select Model

Choose a defensible route before writing code. The output is a modeling argument a reader can follow, not an algorithm label.

Workflow

1. Read the complete problem; list each deliverable, object, datum, unit, objective, constraint and ambiguity.
2. Decompose the problem into subproblems and draw their input/output dependencies.
3. Read references/structure-opportunity-scan.md and inspect shared states, constraints, data, geometry, objectives and solvers. Decide explicitly whether a unified core is justified.
4. Use references/model-catalog.md to shortlist two to four candidates for each subproblem. Compare assumptions, data demand, interpretability, computational cost and what would falsify each candidate.
5. Prefer a route where analytical reasoning exposes structure and numerical methods resolve only the remaining unknowns. Treat reduction, symmetry, monotonicity and event functions as hypotheses requiring evidence.
6. Select a primary route, a simple baseline and a validation route. For threshold search, state the monotonicity evidence or select a non-monotone alternative.
7. State rejected candidates and the reason each fails this problem.

Output

Return a concise 模型选择与结构分析 containing:

• dependency diagram and the unified-core decision;
• per-question objective, variables, units, constraints and assumptions;
• primary, baseline and independent validation methods;
• proposed formulas/outputs, critical events and anticipated figures;
• data requirements, uncertainty/limitation risks and fallback route.

Keep internal decision records machine-readable when a project uses them, but do not mistake a record for an explanation. Do not claim novelty, data availability or expected performance without evidence.

Resources

• references/model-catalog.md — candidate matrix and disqualifying conditions.
• references/structure-opportunity-scan.md — structural reasoning before algorithm choice.

Executable Contract

For repository-managed competition and audit projects, inspect the shared contract registry with python -m scripts.skill_contracts --skill select-model and run this Skill through the local scripts/execute_skill.py with every contracted input and output role. In rapid, return the Output above with a baseline and falsification plan; do not claim the route is formally verified.