gbalabanli/solution-alternatives

Solution Alternatives

Use this skill to produce decision-quality alternatives for hard issues with conflicting constraints.

Workflow

1. Run intake.

• Capture objective, context, hard constraints, soft constraints, preferred risk level, and timeline.
• Ask which optimization mode to use: weighted, lexicographic, or pareto.
• Default to weighted if no mode is selected.

2. Resolve ambiguity with a bounded protocol.

• Ask clarifying questions only when ambiguity can change feasibility or ranking.
• Run up to 2 rounds of clarification.
• In round 1, ask direct questions to fill missing high-impact inputs.
• In round 2, use a different questioning approach from ambiguity_questioning_patterns.md.
• After round 2, stop asking and proceed with explicit assumptions.
• Mark unresolved ambiguity and confidence impact in output.

3. Build and refine the constraint model.

• Classify constraints as hard, soft, derived, or optimization.
• Detect problematic constraints: vague, not measurable, contradictory, or disconnected from the objective.
• Rewrite problematic constraints into measurable forms.
• Add derived constraints when dependency analysis reveals hidden limits.
• Add optimization constraints when objective or utility tradeoff is under-specified.
• Follow constraint_modeling_guide.md.

4. Analyze constraint dependencies and crush risks.

• Identify pairwise dependency status: supports, conflicts, or independent.
• Detect crush risk where optimizing one constraint materially harms another.
• Rate conflict severity and include mitigation notes.

5. Generate alternatives.

• Produce exactly 3 materially different alternatives by default.
• Ensure alternatives are distinct in approach, not minor variants.
• For each alternative, map impacts to the full constraint set.

6. Apply optimization function.

• Use the selected optimization mode rules in optimization_modes.md.
• Always include the optimization function logic in the output.
• In weighted mode, include sensitivity checks and report rank flips.

7. Recommend and define fallback.

• Select one primary recommendation.
• Define explicit switch conditions where runner-up becomes preferred.
• Include next validation step that can confirm or falsify the recommendation quickly.

Output Rules

• Always use the section order in decision_brief_template.md.
• Keep assumptions explicit and scoped to unresolved ambiguity.
• Do not claim certainty when key assumptions remain unvalidated.
• Keep reasoning concise, concrete, and tied to constraint evidence.

Defaults

• Alternatives count: 3
• Ambiguity rounds: 2
• Default optimization mode: weighted

References

• decision_brief_template.md
• constraint_modeling_guide.md
• optimization_modes.md
• ambiguity_questioning_patterns.md