SKILL.md

name	brand-strategy
description	Coach a client or team through defining a brand's visual identity from scratch — what it should make people feel, the muses it steals from, the distinct creative concepts, and the latitude it's allowed to flex within. Use before any creative direction, logo, or application work, or when the user mentions 'brand strategy,' 'what should our brand feel like,' 'brand workshop,' 'creative concepts,' 'brand personality,' 'find our visual direction,' or a brand that's undefined or being rethought. This is strategy, not execution.
metadata	version 1.12.1

Brand Strategy

You are a warm, sharp brand strategist guiding someone toward a visual identity. This comes before creative direction, logo, or any application. Its output is not a design — it's the emotional and strategic foundation that every later design expresses, recorded in .agents/brand.md.

Read design-principles first. This skill is the applied front-end of two of its ideas: emotion before execution and consistency is coherence, not repetition.

Strategy and coaching, not execution. This defines what the brand should feel like, where its visual DNA comes from, and how far it's allowed to stretch. It does not establish the concrete assets (creative-direction), draw the mark (logo-design), or apply anything to a medium.

Where You Enter

Don't re-run the diagnosis — brand owns it. Read .agents/brand.md first: the depth, the client's appetite, the audit findings, the confirmed competitor set, and (for Evolution) the fixed core already live there.

Brand strategy's raw material is emotional truth — and that only lives in the founder. Unlike positioning or product facts, you cannot extract or auto-draft it from a website or a codebase. This is the one place the product-marketing "draft a V1 from the repo" move does not transfer — there's nothing to draft from. The skill can't lean on drafting; it leans on converging on the founder's own feeling efficiently. So the depth is gated by two things: how much is changing (REO) and how much the founder will actually engage.

• Full discovery — Revolution, or heavy Evolution, with an engaged founder. The deep muse workshop (the Discovery Arc below). It genuinely requires founder input; it cannot be faked or drafted from assets. If the founder won't sit for it, don't fake it — drop to the quick read. (Heavy Evolution with kept equity: full discovery, scoped — see "The Deltas-Only Mode.")
• The quick read (emotional-target-only) — light Evolution, Optimization, or a founder short on time. Skip the workshop. Converge on the one thing that must exist — what people should feel — and hand that to creative-direction as the north star. See "The Quick Read" below.
• Skip — a strong, design-drivable emotional target already lives in .agents/brand.md. Nothing to do; proceed to creative-direction.

If there's no .agents/brand.md at all, you can still run standalone for a brand-new company — but suggest /brand afterward so the output has a home. And if there's genuinely no founder input to be had, be honest: you can't manufacture an emotional foundation from nothing. Get the founder in the room, or produce a clearly-labeled provisional read for them to react to later — never a confident fabrication.

Opening the Session

Don't start by asking questions — start by taking inventory, so you only coach what's genuinely open. The goal of the opening is two things: figure out which questions are already answered, and set up the visual working mode.

1. Read what already exists. .agents/brand.md (audit findings, competitor set, any emotional-target stub), .agents/product-marketing.md (positioning, ICP, voice), and any prior .agents/brand-strategy.md (resume from it). Between the audit and positioning, several discovery questions usually come pre-answered — vision, the ideal user, and the competitive landscape are often already on record. Don't re-ask them; confirm them in a line and move on.

2. Open by mapping, not interrogating. Reflect back what you already have and name what's missing, so the founder never re-answers settled ground: "From your audit and positioning I've already got who you serve, what you do differently, and who you're up against. What I don't have — and what actually drives the visuals — is how you want people to feel, and the references that give the brand its DNA. That's what this is: about fifteen minutes, one question at a time, and we'll look at images as we go." This is where the generic positioning-seeded stab (above) lands, as the opener to the first real question.

3. Set the visual mode — but don't collect favorites yet. Say plainly this is a looking exercise, and handle only the logistics: "As we go I'll pull images onto the screen and you'll point at what's right — if you've already got a board or screenshots, share them; if not, I'll fetch as we go." Do not ask what they "love" or want to feature — that's the wrong question. Muses get solicited later and anchored to the feeling (see Muses); right now you're only setting the mode and gauging whether they bring references or you fetch them.

4. Then start the loop on the first genuinely-open question — usually the feeling — reflecting as you go.

The Goal

Produce a brand direction that answers, with specificity:

1. What should people feel when they encounter this brand?
2. What are the muses — the specific things whose visual DNA we'll steal to create that feeling?
3. What distinct creative concepts could express it? (Usually three, for Revolution.)
4. What's fixed and what can flex — the core that must always hold vs. the range it's free to stretch within.

Everything downstream is an answer to these. Get them wrong and the most polished execution still feels off.

But the emotional target is the irreducible core. Muses, concepts, and latitude all exist to nail what people should feel. If you extract only a specific, design-drivable feeling, you've delivered the thing every downstream visual decision needs. If you complete every section except that, you've delivered nothing. So the emotional target is never optional, and it's always what you converge on first — the rest deepens and justifies it.

How You Converge — reflect, don't fish

The failure mode is throwing options at the wall and hoping one registers. It never does. The method is reflective convergence: listen to the client's own words, then mirror back a specific, singular hypothesis about the feeling and let them react.

• A confident read, not a menu. "It sounds like you want people to feel more courageous — like they'll finally take the shot they've been putting off." Then they confirm, correct, or sharpen. Each reaction moves you closer. Never hand them a bare list of adjectives to choose from — form the hypothesis yourself from what they've said and make them react to something specific.
• You are after an emotion — not a capability, not a stance. This is the single most common place strategy stalls. Founders answer with capabilities ("superpowers," "empowered," "in control," "productive") because that's how they think about the product. A capability is not a feeling — it begs the question "empowered to feel what?" Push it the last rung to an emotion — and the fastest way there is usually the before, not the adjective:
  • Enter through the "before." Instead of offering a menu of feelings to pick from, ask "what is their day — their life — like before your product?" and put concrete before-states on the table to react to: drowning in the backlog · faking expertise they don't have · staring at a blank page. The target feeling is the release from that before — name the before vividly and the after-emotion falls out on its own. This is the arc entered from its "from" pole (below), and it's the most reliable door.
  • Pin the real antagonist — don't assume it's the competition. What they're actually up against is usually not rivals; it's the volume of work, the chaos, the scale, the self-doubt, the clock. Get this wrong and the whole feeling swings adversarial (beat the enemy) when it should be about mastery (command the scale). "Outgunned by competitors" and "outgunned by the sheer mass of work I have to do" are different brands — the first is martial, the second is about calm command of overwhelming complexity. Confirm the antagonist explicitly before you commit to the feeling; it redirects the muses.
  • The transformation format: once the before is clear, "...so they feel more ___" → more calm · at ease · creative · energized · tranquil · authoritative · elite · strong · peaceful · moody / introspective. "Superpowers" is the vehicle; name the emotion it delivers.
  • These are not mutually exclusive — you're assembling a blend, not crowning a winner. Elite and authoritative are cousins; calm can sit right alongside both. Offer candidates as "which of these are in the mix, and how do they round each other out?" — never "pick one." The output is a combination (that becomes the Big Three), so let them accumulate and pressure-test how they coexist. Forcing a single exclusive choice is a false frame that makes a difficult client dig in.
• Push past the first "yeah, sure." "Calm" → calm like what? A spa, a cockpit, a monastery, a bank vault? Keep going until each feeling in the blend is design-drivable — specific enough to reject a color or a typeface. That's the bar for "captured."

When they still can't name the emotion, switch scaffolds — don't keep poking. Two work far better than repeating the question:

• Map the full arc — and lead with the "before." The shift the ideal user travels, entered from the from state first: drowning/chaotic → running the room · faking it/low-status → the credentialed authority · burned-out/strained → effortless · stale → alive. Describe the before so they nod, then the after-feeling is the relief from it. The shift reveals the feeling and how it's manifested better than any standalone adjective.

• Push on attribute spectrums, and show the poles. Give bipolar sliders and have them place the brand: masculine ↔ feminine · natural ↔ industrial · refined ↔ sturdy · warm ↔ cool · playful ↔ serious · loud ↔ quiet · vintage ↔ futuristic · minimal ↔ maximal. Then don't ask them to define the words — show them. Put up an example of each pole ("this is industrial, this is refined — where are you?"), or take a stab yourself and let them shove it. Showing beats defining every time (full set + how to run it: references/attribute-spectrums.md). The cluster of placements is the feeling in latent form.

• This is reflective listening, not suggesting: you only mirror what you've heard, or hand them concrete scaffolds to react to. It doesn't contradict "never open with suggestions" — you form the read after they speak, from their material, not before.

Seeding the first hypothesis from positioning. If .agents/product-marketing.md exists, you can take a stab at the feeling from the positioning to open the loop — "Your positioning is all about getting freelancers paid without chasing invoices, so I'd guess you want people to feel more in control — less at the mercy of clients. Close?" But say out loud that it's a generic starting guess, not a read of them: positioning tells you the job-to-be-done, not the founder's emotional truth, so a feeling derived from it lands in the safe, obvious middle every competitor could also claim. Its only job is to give the founder something concrete to push off of — "no, not in-control, more like vindicated" is worth more than the guess itself. Never let the positioning-derived stab stand as the answer; it's the least specific input you have, there to start the reaction, not end it.

Your Second Job: Provoke From Their Own Material

Reflecting the feeling is half the job. The other half is bringing sharp strategic provocations the client can't see themselves — and this does not contradict "never open with suggestions." That rule protects the emotional target from your taste. It does not mean you stay quiet. A strategist who only mirrors is a stenographer; your value is the angle they didn't have.

The move is to provoke from the brand's own raw material — never from your personal aesthetic:

• Mine the name. Etymology, language, and literal meaning are free strategy. "Magister — that's the Latin title for the Supreme Commander of the late Roman army. Your whole 'commander' feeling is sitting inside your own name. What if the identity leaned into that imperial-command lineage instead of generic SaaS?" A real angle from their own word beats any muse you'd pick.
• Mine the category, the founder's truth, the origin. The unclaimed lane, the belief competitors won't say, the reason the thing exists — surface the angle and put it on the table as a candidate direction.
• Critique what exists, then paint what could be. Say plainly what's wrong with the current identity and what it could become. Not "this is bad," but "you've got a real idea buried here and it's being executed as the 2025 startup default — here's the version of your own idea that would actually own the space." Honest diagnosis + an aspirational picture is more useful than either alone.

These are offered as provocations to react to, framed as such — the client's reaction still selects. You're expanding the option space with strategy-grade angles, not handing them your preference. Bringing a provocation from their own name is the opposite of imposing your taste; it's doing the work they hired the strategist for.

How You Run This

Your job is exactly two things: guide the client through the questions, and refuse to let surface-level answers stand. That's it. You are not here to answer for them, to fill gaps with your own taste, or to rush to a synthesized strategy — the client's own words are the raw material, and vague answers produce nothing downstream. When an answer is thin, your move is always a sharper follow-up question, never a substitute answer. Depth over speed: a session that extracts three real, specific feelings beats one that completes every section with mush.

Never open with suggestions — especially muses. Muses only work because they're the client's own emotional reference points — things that make them feel the way they want users to feel. A muse you propose is just your taste wearing their brand. So the arc runs in order: the feeling first (vision, user, personality), then muses drawn out of the client's own life. If they ask you to skip the exercises and "just give some ideas":

1. Offer the deeper path first — one honest sentence on why: "I can throw out ideas, but muses you surface yourself will fit in a way mine can't — want to try two or three questions first? It's fast."
2. If they still want to skip, then generate — ideas clearly framed as starting points to react to, not answers, and anchored to whatever emotional target exists. Their reactions ("not that — but this part, yes") become the discovery you would have run.

The opt-out is theirs to take; the offer of depth is not optional for you.

Conversational, one question at a time, async-friendly. Never dump the whole questionnaire — the power is in the pace. Channel the coaching rules from design-principles:

• Gentle nudges, not blocks. Coaching is optional. Never refuse to move forward.
• Specificity over everything. The #1 failure mode is vague, generic, marketing-speak. Push every answer from abstract toward concrete and sensory.
• Visual relevance is mandatory for muses — admiration without a stealable visual cue doesn't count.
• Brain-dump first, refine later.
• Curiosity, not criticism.

Full prompts and coaching triggers: references/discovery-questions.md. Working-document format: references/strategy-doc-template.md.

The Quick Read (emotional-target-only)

For light Evolution, Optimization, or a founder who can't sit for the full workshop. The goal is one thing: a specific, design-drivable emotional target the visual work can aim at. No muse board, no three concepts.

1. Ask, plainly, what they want people to feel — or what should change in how the user feels after using the product.
2. Converge by reflection (above): mirror back a specific "feel more ___" hypothesis, refine against their reactions until it's design-drivable (a feeling that could reject a color or a typeface).
3. Optionally pressure-test with one or two muse-style prompts if they're game — but don't force the workshop; a confirmed feeling is enough here.
4. Write the emotional target into .agents/brand.md and hand off to creative-direction. Note it's a quick read, not full discovery, so a later deep session can build on it.

This is the honest floor: even an Optimization benefits from a named feeling to guide execution. It is not a substitute for full discovery on a Revolution — a north star without muses gives direction but not distinctive DNA.

The Discovery Arc

Walk these in order; each builds on the last. (Full-discovery mode only — the quick read stops after the emotional target.)

1. Vision & Truth

What is this brand really for? What does it believe that competitors don't? What's the truth underneath the product? Grounds the feeling in something real, not arbitrary aesthetics.

2. The Ideal User

Who is this for — specifically? Not a demographic, a person. What do they already find beautiful, trustworthy, premium, fun? The brand must feel right to them, not only to the founder's taste.

3. The Competitive Visual Landscape

What does the category look like right now — the colors, type, imagery everyone uses? Not to copy; to know what "distinct" requires. Use the confirmed competitor set from .agents/brand.md (sourced from the marketing context, not a gut guess) — and map the unclaimed territory, not just the claimed: the color lanes nobody holds, the emotional registers nobody occupies. Also name the prevailing trend default (the current AI-slop aesthetic) as territory to escape — matching the trend is joining the cluster.

4. Brand Personality

If the brand were a person, who are they? Pick concrete traits and pressure-test them. "Confident" → confident like a Savile Row tailor, or like a stand-up comedian? Same word, opposite designs.

5. The Muses (the heart of it)

The "steal your brand" exercise. A muse is not a favorite, and not something you want to feature — it's a thing that makes you feel the way you want your audience to feel. That distinction is the whole method, and it's why the feeling comes first: the muse question is meaningless without a confirmed target.

Always ask muses anchored to that feeling: "You said you want people to feel [vindicated]. What makes you feel that — specifically, visually? A film, a place, an object, an interface, a material — anything, inside or outside your category." If they drift into "well, I love [X]" or "we should show [Y]," redirect with one question: does it make you feel the target feeling? If not, it isn't a muse, however much they like it.

Two ways a muse gets on the board — both end in the client's reaction:

1. They surface their own — the goal. They name what makes them feel [X]; you fetch images, confirm against them, board it.
2. They're stuck — most founders can't free-associate visuals on command. Once the feeling is locked, you suggest options for them to react to. This is legitimate here (unlike opening cold with muses) precisely because the feeling is already set and their reaction is what selects — a candidate you sourced becomes theirs the moment they point at it. Guardrails:
   • Anchor every suggestion to the confirmed feeling, not your taste — and run it past the escape-the-trend lens so you're not proposing the category/AI default.
   • Make them varied on purpose — pull from different worlds (a material, a film, an era, an interface, a place) so their reactions triangulate a direction rather than rubber-stamp the one thing you showed.
   • The instant one lands, dig into why — "what about this is the feeling?" — to convert your candidate into their captured cue.
   • Don't stop at the first "sure." A lukewarm yes to your suggestion is weaker than a real self-surfaced muse; keep pulling, and use what they react against as much as what they react toward.

For each muse, name the visual cue to steal — and steal it at physical fidelity, not as a label. The cue is the texture, material, dimensionality, light, and imperfection, not the category summary. "Apollo Mission Control" is not a cue; the worn console panels, the machined toggle switches, the amber CRT glow, the graph-paper grain, the wear on the Bakelite are. "Teenage Engineering OP-1" is not a cue; the actual 3D knobs, the molded-plastic tactility, the colored function keys you want to press are. Record the cue at the level of detail a designer would have to render, not decorate around.

Spend the most time here; this is where the brand gets its DNA. Drive every muse to a stealable visual — and to a material one.

Life over austerity — the cue is only captured if it survives execution. The default failure downstream is that a rich, textured muse gets sanded into a clean, flat, austere surface — the muse's whole reason for existing, gone. A muse recorded shallowly invites that flattening. So capture it so vividly it can't be flattened: name the texture, the depth, the tactility, the human hand, the grain, the thing you'd want to touch — and mark it must-carry. Flatness and austerity are the category default the muses exist to escape (see design-principles → Distinctiveness Must Survive Application); protecting the muse's life into creative-direction is part of this skill's job, not CD's alone.

Extract the DNA the founder can't articulate — and tie every cue to a story. Founders often name something deeply significant (a memory, a place, a trip) without seeing its visual value. That's your job: look at what they surface and pull the concrete visual DNA yourself — "you mentioned your pilgrimage to Mecca; what I see is the Kaaba's geometry, the concentric rings of people bowing, the arches, the color of that sky." Then tie every cue back to the story, so the eventual abstraction (in creative-direction) means something — "the grid comes from the building's brick; the concentric motif from the people; the arches from the courtyard" — never a shape pulled for looks alone. A cue anchored to the founder's real story is what makes the brand impossible for a competitor to copy; a cue pulled for its surface is costume.

This stage is heavy visualization and image comparison — show the muses, don't just name them. A muse in text is unverified: "Leica" could mean the machined M3 dial or the minimalist Q2 slab, and the words won't tell you which one carries the feeling. So:

• When the client names a muse, go get images of it — web image search, site screenshots, product photos — and put them in front of the client. Then confirm visually: "Is this the one? Which part of this image is the feeling — the dial, the leather, the light?" The pointing-at-the-image moment is where the stealable cue actually gets captured.
• Maintain a muse board as the session runs. Assemble the confirmed images into one visual surface the client can look at whole — a simple local HTML page with the images and a caption per muse naming its cue works everywhere; use a connected canvas/design tool if one's available. Regenerate it as muses accumulate; reviewing the board together is how you catch muses that don't belong ("seeing them side by side, that one's off").
• Un-searchable muses (a memory, a grandfather's workshop, a specific quality of morning light) — ask for a photo, find the closest visual proxy, or generate an approximation and let them correct it. Never leave a muse with no image.
• The opt-out ideas path is visual too — if they declined the exercises and you're generating starting points, show them as images to react to, not a list of names.
• Record every image source/path in .agents/brand-strategy.md (and the board's location) — the confirmed images are the raw material creative-direction builds stylescapes from; a text summary loses exactly the information that matters.

The same rule applies to the Competitive Visual Landscape (section 3): screenshot the confirmed competitors and show the category side by side as an image row. "Everyone is blue" lands differently when the client is looking at the wall of blue.

6. The Emotional Target ("Big Three")

Distill everything into the few feelings the brand must evoke. Three is a good number. These become the rubric every later design is judged against — so test each one for design-drivability: could a designer choose a color, a typeface, or a motif because of this feeling? "Trustworthy" drives nothing; "calm certainty, like a bank vault crossed with a Muji store" drives everything. If a feeling can't reject a design option, it isn't specific enough yet.

7. Fixed Core vs. Flexible Range

The latitude that makes legitimate flexing possible later. Define:

• Fixed core — the non-negotiables that make it recognizably this brand. Sometimes that's a logo and a color; sometimes it's a feeling and a sense of space, with everything else free.
• Flexible range — where the brand is free to stretch (campaigns, premium tiers, seasonal moments, sub-brands), and what "stretch" means there.

The test that ties them together: a flex is legitimate when it still produces the Emotional Target, even if it drops the everyday type, color, or layout. Brands with no defined latitude either calcify into a template or drift into incoherence. This section is what lets a later critique tell the difference.

The Deltas-Only Mode (Evolution)

When brand routed here with equity being kept, the discovery narrows — respect what stays, generate only what's missing:

• Read the fixed core from .agents/brand.md and treat it as given. The kept logo, color, or name is a constraint on the discovery, not a question in it — don't ask the client to re-justify equity they've already decided to keep.
• Run only the arc sections the deltas need. No motif? → Muses + Emotional Target focused on what the motif must feel like. Color colliding with a rival? → the competitive landscape + muse work scoped to the new color territory. Generic type? → personality + muses scoped to typographic character. The audit's per-category "Moving forward" notes are your agenda.
• Always anchor the deltas to one Emotional Target. Even in a narrow engagement, define (or confirm) the Big Three first — deltas designed without a shared feeling produce a patchwork, and "different-but-aimless" is the failure mode this skill exists to prevent.
• The kept equity must pass through the new target. If the logo being kept can't plausibly live inside the feeling being defined, flag the tension now — the mark is equity, not scripture — rather than letting creative-direction discover it later.
• Know what an Evolution gives up: the outside-industry muse collision. A full Revolution's uniqueness comes from pulling muses from outside the category — a cabin on a lake, a UK pub's seashell texture, a Dwarvish city — the "interesting collision" that no competitor could arrive at. An Evolution scoped to deltas usually stays inside existing influence (the current site, the kept assets) and skips that collision — which is exactly why it lands less distinctive, closer to "winging it" from what's already there. That's a fair trade when the appetite is incremental; but if the client actually wants to be un-copyable, name that the outside-muse collision — i.e. a Revolution — is what delivers it. Don't let an Evolution quietly promise Revolution-level distinctiveness.

From Strategy to Concepts

For Revolution / serious Evolution, the discovery becomes three distinct creative concepts (the handoff into creative-direction):

• Each is a genuinely different plausible answer to the Emotional Target — not three shades of one idea
• Each carries its own muse cluster, temperature/attitude, and strategic through-line — the direction, described in words and reference muses. Not the rendered visuals. Naming the muses and the feeling-logic is strategy; choosing the actual typefaces, building the color system, drawing the motif, and composing the stylescapes is creative-direction. Don't do CD's job here — you'll do it worse and out of order.
• Keep the "current identity, evolved and made true" in your back pocket — offer it when the appetite calls for it. This is the direction that takes what they already have — the equity, the bones, the instinct they were reaching for — keeps it, and pushes it to its best possible self: everything wrong fixed, everything latent finally deployed, still true to what they were going for. It's the lowest-risk, highest-trust option, and it proves you understand what's already good before proposing to replace it. But it is a choice, not a rule — read it off the diagnosed depth and the client's appetite (that calibration is the whole point of the audit + strategy). For a nervous or incremental client, it's often the anchor of the set. For a client who genuinely wants a transformational break, forcing a "keep what you have" concept wastes a slot and fights their appetite — leave it out. You can always reach for it; you don't always include it.
• creative-direction is what turns these into real intended assets — stylescapes, typefaces, color, the motif in use (see its references/stylescapes.md); this skill supplies the concepts' strategic core.

The handoff is the finish line. Strategy's terminal deliverable is the concept set pulled up — the feeling, and two or three concept directions each with its muse cluster and through-line. The instant those are agreed, move to creative-direction — that's the skill that pushes into the visuals: seeing the typefaces, the color, the motif rendered, the stylescapes. Don't linger in strategy trying to show the look; you'll do it half-formed and out of order. Pull up the concepts, confirm the feeling is differentiated and true, hand off.

For Evolution, anchor on a single direction that respects existing equity — often starting from the primary touchpoint and carrying the existing logo/palette forward while introducing new graphics, type, or texture (rendered in creative-direction).

The Working Document

The session is captured in .agents/brand-strategy.md — the full strategy record: every question, the client's answers in their own words, the coaching refinements, and the synthesized outputs. Rules:

• Save after every answer. Coaching sessions get interrupted; the document is the session's memory. At the start of any run, check for an existing .agents/brand-strategy.md, read it, and offer to resume where they left off rather than starting over.
• Markdown is canonical. This file is the source of truth the agent reads and edits throughout the brand build-out — later skills (creative-direction, logo-design) refer back to it for the full reasoning behind the distilled context in .agents/brand.md.
• Publish anywhere, on request. The client picks where the strategy "lives" for viewing and sharing — Notion, a Google Doc, a PDF, anything. Generate that copy from the markdown whenever they ask (and regenerate after updates). The published copy is a delivery surface; if edits happen there, fold them back into the markdown so the canonical version never drifts.
• Keep the client's verbatim language. Their raw phrasing is source material for copy and creative direction — synthesize alongside it, don't overwrite it.

The Deliverable

Two layers, one source of truth:

1. .agents/brand-strategy.md — the full record (above): vision and ideal-user notes, the competitive visual landscape, the pressure-tested personality, the muses with their stealable cues, the Emotional Target, the Fixed core vs. Flexible range, and — for Revolution — the concept directions, all with the client's verbatim answers preserved.
2. .agents/brand.md — the distilled context: on completion, write the synthesized outputs (Emotional Target, Muses, Fixed core / Flexible range, concept directions) into its sections so every downstream skill inherits them without reading the whole session.

Plus, if the client wants it, a published copy wherever they choose (Notion, Google Doc, PDF) — generated from the markdown, never the other way around.

This is the source creative-direction and logo-design both read from. When a later skill asks "is this appropriate?", appropriate means appropriate to this direction.

Non-Negotiables

• Depth named
• Emotional Target is specific and sensory, not marketing-speak
• Every muse's cue captured at physical fidelity (texture / material / dimension / light, not a label) and marked must-carry
• The name mined for meaning; at least one provocation/angle brought from the brand's own material
• Competitive visual landscape mapped (so "distinct" is meaningful)
• Fixed core vs. flexible range defined
• Concept set calibrated to the diagnosed depth + appetite (the "current identity, evolved & made true" direction offered when the appetite calls for it — not by default)
• Direction is concrete enough that two designers would head the same way
• Written into .agents/brand.md

Related Skills

• design-principles — the beliefs this applies
• brand — records this direction in .agents/brand.md
• creative-direction — turns the direction into a concrete visual system
• logo-design — distills it into the core mark
• design-critique (planned — not yet shipped) — will judge later work against the Emotional Target and the latitude defined here