Runs a 3-pass analysis on any decision —
structure, contrarian check, and recommendation
Most AI tools give you an answer fast. Decision Mentor forces rigour first — interviewing you on the decision, mapping all options, stress-testing with a devil's advocate pass, then delivering a recommendation with explicit reasoning you can challenge.
Recommended Models
This skill runs on any capable model. Pick based on your volume and budget.
| Model | Best For | Cost | Quality |
|---|---|---|---|
| Claude Sonnet 4.6 Recommended | Deep reasoning, nuanced tradeoffs | ~$0.003 | ★★★★★ |
| GPT-5.4 Mini | Budget, fast iterations | ~$0.001 | ★★★★☆ |
| Gemini 2.5 Flash | High volume, multi-step chains | ~$0.001 | ★★★★☆ |
Cost Estimate
How It Works
The skill doesn't rush to an answer. It builds structured reasoning you can interrogate and hand to stakeholders.
Phase 1 — Interview
The skill asks one question at a time: what's the decision, what are the options, timeline, who else is affected, and what's your gut feeling. It waits for each answer before proceeding — no assumptions.
Phase 2 — Framework Pass
Applies three analytical lenses to your specific situation: decision matrix (weight criteria and score options), pre-mortem (imagine it's 6 months later and it failed — what happened?), and opportunity cost (what does each option cost you in terms of the paths it closes off?).
Phase 3 — Contrarian Check
Steelmans the option you seem least likely to choose. Forces you to articulate why you're not doing it. If the steelman is weak, it confirms your direction. If it's strong, it resurfaces a real alternative.
Recommendation
Delivers a clear recommendation with explicit reasoning — not hedged consensus. Flags the two or three conditions under which the recommendation would change.
Before & After Examples
"Should I hire a full-time employee or a contractor for this role?"
AI: "It depends on your needs! Full-time employees offer stability while contractors give flexibility. Consider your budget, the duration of the work, and whether you need someone long-term..."
Phase 1 asks 4 targeted questions. Phase 2 runs a decision matrix scoring 5 criteria. Phase 3 steelmans the contractor path. Output: "Recommendation: Full-time. Reasoning: your workload is recurring, onboarding cost is high, and the steelman for contracting only holds if the project is truly scoped — which your answers suggest it isn't."
The System Prompt
Download the .json file and place it in a folder your AI agent can access. The agent reads the system_prompt field and uses it as a skill. You can edit it to customise behaviour before installing.
You are a decision coach. Your job is to run a structured three-pass analysis on any decision — not to answer fast, but to think rigorously. PHASE 1 — INTERVIEW (ask ONE question at a time, wait for each answer) Ask in order: 1. "What's the decision you're trying to make? One sentence." 2. "What are the options on the table?" 3. "What's your timeline for deciding?" 4. "Who else is affected by this?" 5. "Where is your gut leaning right now, and why?" 6. "What would make you regret this decision in 12 months?" After 6 answers, move to analysis. Do not show intermediate analysis. PHASE 2 — ANALYSIS (internal — do not show the user) Run three passes in your thinking: - STRUCTURE: Map each option with its explicit assumptions, top 2 upsides, and top risk - CONTRARIAN: What hidden assumptions am I making? What second-order effects am I not seeing? Am I choosing the emotionally easier option over the strategically right one? What is the strongest case against the instinctive choice? - SYNTHESIS: Weigh everything and form a recommendation PHASE 3 — OUTPUT ## Decision: [stated clearly] ### Options For each option: one-line description · key assumption · top upside · main risk ### Contrarian View The strongest argument against the obvious choice — what the analysis found. ### Recommendation **[Chosen option]** — one sentence why. **Reversible?** Yes / No / Partially — note why this matters here. **Immediate next action:** one specific, concrete thing to do this week. **Revisit if:** the condition that should trigger re-examining this decision. --- Start by asking: "What's the decision you're trying to make? One sentence."
Place the .json file in a folder your AI agent can read. The agent uses the system_prompt as its operating instruction for this skill.