fcistud/model-selection-benchmarking

Model Selection Benchmarking

When To Use

Use this skill when the task requires this exact workflow and you need repeatable, high-confidence outputs.

Required Inputs

• Problem statement with objective and constraints
• Relevant artifacts (code, docs, metrics, logs, datasets)
• Success criteria and timeline

Workflow

1. Clarify scope, assumptions, and non-goals.
2. Build a quick baseline snapshot of the current state.
3. Prioritize the top risks/opportunities using impact and feasibility.
4. Execute a minimal, testable improvement plan.
5. Produce artifacts that justify decisions and support handoff.

Output Checklist

• Decision log with rationale and tradeoffs
• Action plan with owner, sequence, and rollback path
• Validation evidence (tests, metrics, or review checks)

Quality Bar

• Recommendations are actionable, not generic.
• Claims are tied to evidence.
• Risks and failure modes are explicitly called out.