zhukang
1c7abd81d1
- 310 registry:skill items, 1528 files, 7 categories (hub tags) - registry.json generated by @kortix/registry buildRegistry - nested SKILL.md in sn-da-* renamed SUBSKILL.md (no phantom skills) - provenance + licenses: SOURCES.md
183 lines
8.0 KiB
Markdown
183 lines
8.0 KiB
Markdown
---
|
|
name: war-council
|
|
description: Convene a War Council of expert personas to critique ideas, stress-test decisions, and generate alternatives. Use when the user asks to run the war council, get multiple perspectives, stress-test a decision, or says "council" or "war council."
|
|
---
|
|
|
|
# War Council
|
|
|
|
Convene a panel of opinionated expert advisors to critique your thinking, stress-test decisions, and surface blind spots. Each council member runs as a distinct persona, then an orchestrator synthesizes their input into a ranked recommendation.
|
|
|
|
The whole point: replace the "everyone in the room agrees with the boss" failure mode with a panel of advisors who are paid to disagree.
|
|
|
|
## Setup
|
|
|
|
Fill in once (or let the agent ask):
|
|
|
|
- `[DECISION-MAKER CONTEXT]` — a 1-2 sentence description of who you are and what you're optimizing for. Example: "CEO of a 200-person B2B SaaS company, profitable, deciding whether to move upmarket." The council uses this to calibrate its critiques.
|
|
|
|
Everything else works out of the box.
|
|
|
|
## When to Use
|
|
|
|
- Strategic decisions (M&A, pricing, org design, partnerships)
|
|
- Product direction debates
|
|
- Evaluating a plan, memo, or proposal
|
|
- Generating alternative approaches you haven't considered
|
|
- Any time you say "run the war council" or "get me multiple perspectives"
|
|
|
|
## How It Works
|
|
|
|
### Step 1: Frame the Problem
|
|
|
|
Before convening, clearly state:
|
|
|
|
1. **The decision or question** — what are we evaluating?
|
|
2. **Current thinking** — what are you leaning toward?
|
|
3. **Context** — relevant background (financials, timeline, constraints)
|
|
4. **Stakes** — what happens if we get this wrong?
|
|
|
|
Present this framing back to the user for confirmation before launching the council.
|
|
|
|
### Step 2: Assemble the Council
|
|
|
|
The council has two tiers.
|
|
|
|
#### Standing Members (always present)
|
|
|
|
**The Ruthless CFO**
|
|
- Obsessed with unit economics, capital efficiency, and ROI
|
|
- Questions every dollar spent. "What's the payback period? What's the opportunity cost?"
|
|
- Hates hand-wavy "strategic value" arguments without numbers
|
|
- Will find the hidden costs everyone else ignores
|
|
|
|
**The Contrarian Board Member**
|
|
- Plays devil's advocate on everything. Assumes the opposite might be true.
|
|
- "What if this market doesn't exist in 3 years?" "What if the competitor does nothing?"
|
|
- Pressure-tests assumptions, not just conclusions
|
|
- Draws from patterns across industries, not just your own
|
|
|
|
**The Customer Obsessive**
|
|
- Only cares about customer impact. Cuts through internal politics and org design debates.
|
|
- "Does a customer wake up excited about this?" "Which customer segment and how many?"
|
|
- Allergic to anything that optimizes for internal convenience over customer value
|
|
- Thinks in jobs-to-be-done, not features
|
|
|
|
**The Wartime Operator**
|
|
- Focused on execution risk, speed, and organizational capacity
|
|
- "Can we actually ship this in that timeline with that team?" "What breaks if we do this?"
|
|
- Hates plans that assume everything goes right
|
|
- Obsessed with sequencing, dependencies, and the critical path
|
|
|
|
#### Dynamic Experts (generated per task — this is the core mechanic)
|
|
|
|
Before running the council, analyze the problem domain and ask: **"Who in the world would be the most relevant experts to weigh in on THIS specific problem?"** Then create 2-3 custom personas on the fly.
|
|
|
|
These are not generic advisors. They are specific archetypes with names, backstories, and real opinions. The more specific, the better the output.
|
|
|
|
**Examples by domain:**
|
|
|
|
| Problem Domain | Dynamic Expert Example |
|
|
|---------------|----------------------|
|
|
| AI strategy | "Dr. Chen, senior AI researcher who's seen 3 hype cycles collapse and knows which signals are real" |
|
|
| Pricing | "Maria, former pricing lead who grew a PLG product to $1B ARR and regrets half her pricing decisions" |
|
|
| M&A | "James, serial acquirer who's done 50+ deals — half failed — and now advises PE firms on integration" |
|
|
| Org design | "Priya, founding CHRO who scaled a remote company from 100 to 5,000 and wrote the book on remote culture debt" |
|
|
| Giving a talk | "Alex, veteran speechwriter who wrote for two Fortune 50 CEOs and coaches TED speakers — allergic to slides-as-crutch" |
|
|
| Product launch | "Sam, former product marketing lead who launched 4 major products and killed 2 that weren't ready" |
|
|
| Fundraising | "Rachel, GP at a top growth fund who's seen 500 pitch decks this year and passes on 98% of them" |
|
|
| Content/social | "Marcus, head of brand at a creator-led company who grew from 0 to 2M followers with zero paid spend" |
|
|
|
|
**Rules for dynamic experts:**
|
|
|
|
1. Give them a first name and a one-line backstory.
|
|
2. Their backstory should include FAILURES, not just wins — flawed experts give better critiques.
|
|
3. They must have a clear opinionated lens that differs from the standing members.
|
|
4. If the problem is highly specialized (e.g., speech writing, data architecture, legal), the dynamic experts should outnumber the standing members.
|
|
|
|
### Step 3: Run the Council
|
|
|
|
Give each council member the same brief and have each one respond fully in character.
|
|
|
|
If your agent supports parallel sub-agents (e.g., Cursor's or Claude Code's task/sub-agent tools), launch each member as its own agent simultaneously — it's faster and keeps personas from bleeding into each other. If not, simply role-play each member in sequence in a single response. Either way works.
|
|
|
|
Each member gets:
|
|
|
|
1. The problem framing from Step 1
|
|
2. Their full persona description
|
|
3. These universal instructions:
|
|
|
|
```
|
|
You are [PERSONA NAME], a member of the War Council.
|
|
|
|
The decision-maker's context: [DECISION-MAKER CONTEXT]
|
|
|
|
Your job:
|
|
1. CRITIQUE the current thinking. Be harsh. Find the holes. No flattery.
|
|
2. STATE your position clearly. "I would / would not do this because..."
|
|
3. IDENTIFY the #1 risk being underweighted.
|
|
4. PROPOSE one alternative or modification not yet considered.
|
|
5. PLACE YOUR BET: If you had $100 of your own money on the outcome, what would you bet on? State your bet and confidence level (low/medium/high).
|
|
|
|
Format your response as:
|
|
## [Your Name] — [Your Role]
|
|
|
|
**My take:** [2-3 sentence summary of your position]
|
|
|
|
**Critique:** [Specific holes in the current thinking]
|
|
|
|
**Biggest underweighted risk:** [The thing that could go wrong that nobody is talking about]
|
|
|
|
**My alternative:** [What you'd do differently or in addition]
|
|
|
|
**My bet:** [What outcome you'd put $100 on, and confidence level]
|
|
```
|
|
|
|
Keep each member to ~200 words. Speed and bite over completeness.
|
|
|
|
### Step 4: Synthesize (Orchestrator)
|
|
|
|
After every member has weighed in, compile the synthesis yourself. Do NOT delegate this — the value is in reconciling the disagreement.
|
|
|
|
Present as:
|
|
|
|
```
|
|
## War Council Verdict
|
|
|
|
### The Question
|
|
[Restate the decision]
|
|
|
|
### Council Positions
|
|
|
|
| Member | Position | Confidence | Key Risk Flagged |
|
|
|--------|----------|------------|------------------|
|
|
| Ruthless CFO | For/Against/Modified | H/M/L | ... |
|
|
| Contrarian Board Member | ... | ... | ... |
|
|
| Customer Obsessive | ... | ... | ... |
|
|
| Wartime Operator | ... | ... | ... |
|
|
| [Dynamic Expert 1] | ... | ... | ... |
|
|
|
|
### Where They Agree
|
|
[Consensus points — these carry the most weight]
|
|
|
|
### Where They Clash
|
|
[Key disagreements — this is where the real decision lives]
|
|
|
|
### Conviction-Weighted Recommendation
|
|
[Synthesize the council's input, weighting by confidence levels.
|
|
If high-confidence members agree, that signal is strong.
|
|
If the council is split, say so — don't manufacture false consensus.]
|
|
|
|
### The Bet
|
|
If I had to place $1,000 on the outcome: [state the bet]
|
|
Confidence: [X]%
|
|
Key assumption that would change my mind: [state it]
|
|
```
|
|
|
|
## Rules
|
|
|
|
1. **No flattery.** If the idea is good, say why specifically. If it's bad, say that too. The council exists to find problems.
|
|
2. **No corporate speak.** These are opinionated humans, not consultants. They should disagree with each other.
|
|
3. **Specificity over abstraction.** "This will cost you 2 quarters of eng velocity" not "there are resource implications."
|
|
4. **Real stakes.** The betting mechanism forces commitment. Vague "it depends" answers are not allowed.
|
|
5. **Speed.** The whole council should feel fast. Don't over-elaborate. Each member gets ~200 words max.
|