cosmicstack-labs/decision-matrix
A structured reasoning assistant that helps you make clear, confident decisions using proven frameworks — Pros/Cons, Decision Matrix (Pugh Matrix), Weighted Scoring, Pre-mortem, and Opportunity Cost Analysis.
npx skills add https://github.com/cosmicstack-labs/mercury-agent-skills --skill decision-matrix
Takes the guesswork out of tough decisions by applying structured frameworks. Instead of spinning in circles with pros/cons lists, you get a repeatable system for comparing options, weighting what matters, and reaching a conclusion you can explain and defend.
Best for: Quick decisions with low-to-moderate stakes
| Step | Action |
|------|--------|
| 1 | Draw two columns: PROS and CONS |
| 2 | List every reason for and against — no filtering |
| 3 | Weigh each item (not all pros are equal). Assign +1 to +5 for pros, -1 to -5 for cons |
| 4 | Sum the scores. A net positive suggests "yes." Consider emotional weight too |
Guardrail: Pros/cons alone miss hidden assumptions. Always follow with: "What am I not considering?"
Best for: Comparing multiple options against multiple criteria
| Criteria | Weight (1-5) | Option A | Option B | Option C |
|------------------------|-------------|----------|----------|----------|
| Cost | 4 | 8/10 | 6/10 | 9/10 |
| Time to Market | 3 | 7/10 | 9/10 | 5/10 |
| Strategic Fit | 5 | 9/10 | 4/10 | 7/10 |
| Team Capacity | 2 | 6/10 | 8/10 | 4/10 |
| **Weighted Total** | | 113 | 92 | 103 |
Steps:
Best for: High-stakes decisions where risk mitigation is critical
> "It's 12 months from now and our decision has failed spectacularly. How did it happen?"
| Step | Technique |
|------|-----------|
| 1 | Assume the decision was made and led to disaster |
| 2 | Fast-forward and write the "post-mortem" — what went wrong? |
| 3 | Generate 5-10 plausible failure modes |
| 4 | For each failure, ask: "What could prevent this?" |
| 5 | Incorporate those safeguards into the decision |
Why it works: Humans are loss-averse. Imagining failure activates risk-awareness that simple pros/cons don't reach.
Best for: Deciding between two good options (where saying yes to A means saying no to B)
| Frame | Question |
|-------|----------|
| Cost of yes | What do I give up by choosing this? |
| Cost of no | What do I give up by not choosing this? |
| Regret test | If I look back in 5 years, which "no" would I regret more? |
| Opportunity comparison | If Option A didn't exist, would I choose Option B? |
Heuristic: If saying "no" to Option A doesn't feel like a loss, don't choose it.
Best for: Prioritizing many options quickly (features, ideas, experiments)
| Criterion | Scale | Question |
|-----------|-------|----------|
| Impact | 1-10 | How significant will the result be if successful? |
| Confidence | 1-10 | How sure are we about the expected outcome? |
| Ease | 1-10 | How easy/simple is this to execute? |
Formula: ICE Score = Impact × Confidence × Ease
Sort by score. Work on the highest first. Re-score when new data emerges.
Best for: Emotional or high-stakes personal decisions
| Time Horizon | Question |
|-------------|----------|
| 10 minutes | How will I feel about this decision in 10 minutes? |
| 10 months | How will I feel about it in 10 months? |
| 10 years | How will I feel about it in 10 years? |
Purpose: Shifts perspective from short-term emotion to long-term impact. If all three horizons align — it's an easy call. If they conflict, the 10-year horizon should usually win.
| Phrase | Action |
|--------|--------|
| "Help me decide between..." | Starts a structured comparison of options |
| "Pros and cons of..." | Generates a weighted pros/cons table |
| "Should I [X] or [Y]?" | Runs a decision matrix or opportunity cost analysis |
| "What am I not considering?" | Surfaces blind spots and hidden assumptions |
| "Run a pre-mortem on..." | Scenarios worst-case outcomes to de-risk the decision |
| "Prioritize these for me..." | Uses ICE or weighted scoring to rank options |
| "Help me think this through..." | Combines frameworks layered for clarity |
A fuzzy question gets a fuzzy answer. Be specific:
| Decision Type | Recommended Framework |
|---------------|---------------------|
| Low stakes, 2 options | Pros & Cons (weighted) |
| Multiple options, many criteria | Weighted Decision Matrix |
| High risk, irreversible | Pre-mortem |
| Scarcity (time/money focus) | Opportunity Cost Frame |
| Prioritizing a long list | ICE Score |
| Emotional/personal | 10/10/10 Rule |
Gather:
Run the framework step by step. Document scores, weights, and reasoning.
| Bias | Mitigation |
|------|-----------|
| Confirmation bias | Actively list reasons *against* your preferred option first |
| Recency bias | Consider decisions from 6+ months ago — does this feel different? |
| Sunk cost | "If I had no prior investment in this, would I still choose it?" |
| Status quo bias | "If this weren't the default, would I pick it?" |
After the decision plays out, revisit your framework. Did your weights reflect reality? Did you miss a criterion? Retrospect improves future decisions.
> Input: "Should I take Client A ($5k, urgent, boring) or Client B ($3k, flexible, exciting project)?"
>
> Process: Weighted Decision Matrix
>
> | Criteria | Weight | Client A | Client B |
> |----------|--------|----------|----------|
> | Income | 4 | 9 (36) | 5 (20) |
> | Enjoyment | 3 | 3 (9) | 9 (27) |
> | Time Pressure | 2 | 3 (6) | 9 (18) |
> | Portfolio Value | 4 | 4 (16) | 9 (36) |
> | Total | | 67 | 101 |
>
> Result: Client B wins despite lower pay, because portfolio value and enjoyment outweigh the income gap.
> Input: "Should I prioritize building a mobile app or improving onboarding?"
>
> Process: ICE + Pre-mortem
>
> ICE:
> - Mobile App: Impact 8, Confidence 4, Ease 2 → ICE = 64
> - Onboarding: Impact 6, Confidence 8, Ease 8 → ICE = 384
>
> Pre-mortem on mobile app decision: "We built the app but no one used it because onboarding was broken." → Clear signal to fix onboarding first.
Take cosmicstack-labs/decision-matrix from the repository into ~/.claude/skills for personal
use, or into .claude/skills inside a project.
The agent identifies a skill by the name field in its header. Two skills with the
same name cannot sit side by side — one of them will be ignored.