Council

The pattern in one phrase: many lenses on one thing.

1. What it does

It runs your question past a panel of distinct expert "characters," each one judging it from their own angle, then merges their takes into a single verdict with ranked next steps.

2. The problem it solves

When you ask one smart helper to review your work, you get one point of view — and one point of view has blind spots. A copywriter loves your headline but never notices the pricing is wrong. A designer fixes the look but misses that nobody can find the buy button. You walk away feeling reviewed, but the thing that actually sinks you was never in that person's lane. It's like getting your car checked by only the guy who does tires: he says you're good, and three weeks later the transmission goes. You needed a whole crew looking at it, each at their own part.

3. The core pattern (the part you steal)

Many lenses on one thing.

The transferable idea: take a single input, and deliberately look at it through several *different fixed viewpoints* before you decide anything. Each viewpoint is a "character" with a narrow obsession — one only cares about money, one only cares about how it makes people feel, one only cares about whether it's bold enough to stand out. None of them is the whole answer. The power is in the *spread*: where one character is blind, another is staring right at the problem.

Think of a hospital tumor board. One patient scan goes in front of a surgeon, a radiologist, and an oncologist at the same table. The surgeon sees what they'd cut, the radiologist sees what the image really shows, the oncologist sees the drug plan. No single doctor is overruled or trusted blindly — they each speak from their specialty, then the room agrees on one treatment plan. You catch things no solo doctor would, because three trained eyes were forced to look at the *same* picture from three angles.

That's the whole pattern, and it has nothing to do with branding specifically. Code reviews, hiring decisions, big purchases, business pitches — anything you'd want a "second opinion" on actually wants a *fifth* opinion from people who care about different things. The skill just makes that panel show up on command.

4. How the skill is actually built

Look at the real file: ~/.claude/skills/brand-council/SKILL.md. Its shape:

• Frontmatter (the top block): name plus a description packed with trigger phrases like *"brand council," "review my brand," "does this brand work?"* Those phrases are the doorbell — they tell the agent when to wake this panel up instead of just answering as itself.
• The roster (the heart of it): the body is mostly a list of *characters*. Each one gets the same four fields — Who they are (their real track record, so the voice has weight), their Lens (the one thing they obsess over), the questions they always ask, and their Voice (so a money-person and a culture-person sound nothing alike). This repeated structure is the engine: same template, different obsession, over and over.
• The process steps: gather context first (what is this, who's it for, what stage), then each character grades the same input on a simple scale and names their single biggest issue, then a synthesis step collapses all of it into one scorecard plus a ranked "fix this first" list. The fan-out (many takes) always ends in a fan-in (one verdict) — otherwise you've just got noise.
• Helper detail: each character points to a fuller "advisor" file of its own, so the panel can borrow a deeper version of any single voice when one lens needs to go deep. The council is the quick panel; the advisors are the specialists on call.

The trick that makes it work: the characters are *fixed and different on purpose*. If they all thought alike, you'd just be asking the same person five times.

5. How YOU could build your own

You don't need famous names. You need a few sharply different jobs:

1. Pick 3-5 viewpoints that would fight. For a sales page that might be: the money person, the "does this feel trustworthy" person, the "is this clear in five seconds" person, the "would I share this" person. The more they'd disagree, the better the coverage.
2. Give each one a one-line obsession and a couple of questions they always ask. That's all a "character" really is — a narrow lens plus a habit of asking the same pointed questions every time.
3. Force a fan-in. After every character speaks, add one final step that says: combine these into a single verdict and a top-3 list of what to fix first, in order. Without this you get five opinions and no decision.
4. Add trigger phrases to the description so the agent convenes the panel on its own when you say "review this" — instead of you having to summon each voice by hand.

That's a working council. Famous personas and grading scales are flavor you can add later.

6. When to use it — and when NOT to

• Use it when: the decision has several dimensions that one expert can't all cover (a brand, a pitch, a design, a hire), the cost of a blind spot is high, or you keep getting "looks good" reviews and then getting burned by something nobody flagged.
• Skip it when: the question has one right answer or one clear owner. A factual lookup, a quick yes/no, or "fix this typo" doesn't need a panel — five viewpoints on a one-lane question is just five people agreeing slowly. Don't convene a tumor board to check a paper cut.

7. The stripped-down version

This is a complete, working council skill. Swap the four roles for whatever your decision needs:

---
name: council
description: Review one thing through several expert viewpoints, then give one verdict. Use when I say "review this," "council," or "get me a second opinion."
---

# Council

When I bring you something to review, judge it through these four characters, one at a time:

- MONEY: only cares "will someone pay, and is the path to revenue clear?"
- TRUST: only cares "does this feel credible, or is something off?"
- CLARITY: only cares "can a stranger get it in five seconds?"
- SHARE: only cares "would anyone send this to a friend, and why?"

For each: give a grade (A-F), their single biggest issue, and one quick fix.

Then SYNTHESIZE: one overall verdict + a ranked top-3 list of what to fix first.