Ml Problem Framing

Overview

Use this skill to define an ML problem that supports a real product decision with measurable outcomes.

Scope Boundaries

• Use this skill when the task matches the trigger condition described in description.
• Do not use this skill when the primary task falls outside this skill’s domain.

Shared References

• Objective and labeling rules:
  • references/objective-and-labeling-rules.md

Templates And Assets

• Problem framing template:
  • assets/problem-framing-template.md

Inputs To Gather

• Business decision to support and value target.
• Candidate prediction target and labeling source.
• Risk constraints (fairness, latency, compliance).
• Baseline process and non-ML alternatives.

Deliverables

• Framed ML objective with explicit non-goals.
• Label definition and prediction unit.
• Success metrics and decision thresholds.
• Risks and assumptions log.

Workflow

1. Define decision context with assets/problem-framing-template.md.
2. Validate objective/label choices using references/objective-and-labeling-rules.md.
3. Align metric choices to business and user outcomes.
4. Document assumptions, constraints, and alternatives.
5. Publish go/no-go framing decision.

Quality Standard

• Objective is measurable and decision-relevant.
• Label definition is leakage-safe and reproducible.
• Metrics and thresholds are operationally actionable.

Failure Conditions

• Stop when objective does not map to a concrete decision.
• Stop when label quality/timing cannot be validated.
• Escalate when framing risks exceed policy tolerance.