agentleFS
LittlpinguinLittlpinguin/marketing-cockpit-template/01-brand/CLAUDE.md

marketing-cockpit-template / 01-brand

CLAUDE.md9 starsChanged 13 days agoMITView on GitHub

This folder is the single source of truth for identity, voice, messaging, and visual identity. It produces nothing; it is consulted by every other role (02-strategy/ through 09-seo/) before any content is created. Absolute rule. When 01-brand/ and another folder disagree, 01-brand/ wins. Surface the conflict to the user before acting. Published examples. Re-reading the 3 to 5 most recent pieces of the target channel before drafting stays the normal procedure, for anti-repetition as for tone calibration: it is the…

The file

We split this CLAUDE.md into the rules it gives an agent

6 rules. About 0 are common, 0 are distinctive, and 6 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈0

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 6

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Absolute rule: When 01-brand/ and another folder disagree, 01-brand/ wins.

    line 7
  2. Very specific:

    Read the files directly — this folder is small enough that file reads are cheap.

    line 90
  3. Very specific:

    Before any content ships, the producing role must pass this filter — this is what the brand-check skill enforces.

    line 107
  4. Very specific:

    If a rule is missing: do not invent. Escalate to {{COMPANY_MAIN_CONTACT}}.

    line 120
  5. Very specific:

    If a word joins the banned list: add it to voice.md, then to [forbidden-words.charter] of scripts/lint-brand.toml.

    line 122
  6. Very specific:

    If a number becomes outdated: update messaging-framework.md and brand-platform.md together.

    line 123

≈ “Common” is an estimate: at least 1 in 5 of the 27,641 CLAUDE.md files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
CLAUDE.md

01-brand — keeper of the {{COMPANY_NAME}} brand

Role

This folder is the single source of truth for identity, voice, messaging, and visual identity. It produces nothing; it is consulted by every other role (02-strategy/ through 09-seo/) before any content is created.

Absolute rule. When 01-brand/ and another folder disagree, 01-brand/ wins. Surface the conflict to the user before acting.

When to consult this folder

You need to…Read in priority
Write a post, email, page, event scriptvoice.md + messaging-framework.md
Target a specific audiencepersonas.md
Pick a number or a proof pointmessaging-framework.md (key numbers section)
Use a tagline, quote, or signature phrasevoice.md (signature phrases)
Design a visual, pick a color or fontstyle-guide.md (intent) + tokens.json (exact values)
Check a draft for AI tellsanti-ai-writing-style.md
Know what the brand already refusedexemples-rejetes.md
Publish something AI-generateddivulgation-ia.md
Publish a third-party logo, a photo of a person, a fontdroits.md
Understand full strategic contextbrand-platform.md
Identify an internal stakeholderstakeholders.md
Reuse a visual assetassets/

Published examples. Re-reading the 3 to 5 most recent pieces of the target channel before drafting stays the normal procedure, for anti-repetition as for tone calibration: it is the most reliable material an agent has. What is forbidden is something else: writing a specific deliverable into a skill as the model to imitate, which makes every later production converge on it. Quoting a published piece in a discussion, or using it to calibrate: allowed. Freezing it as a reference inside a skill: not allowed.

Task router

Entry point of any production: read what the row says, in that order, and nothing more. A file already read in the session is not reloaded.

DeliverableFiles to read, in order
Text post or emailvoice.md → messaging-framework.md (key numbers) → anti-ai-writing-style.md → exemples-rejetes.md
Visualtokens.json → assets/ catalogue → divulgation-ia.md → droits.md if a third-party logo or a person appears
Deck or carouselthe two rows above, text then visual, plus style-guide.md (type scale)
Web pagevoice.md → messaging-framework.md → style-guide.md → tokens.json

File inventory

These files are created by /brand-discover. If any are missing or empty, the wizard has not run to completion or the user deliberately skipped the section (see _gaps.md).

FileContentWhen to use
voice.mdVoice position, preferred/banned vocabulary, typography rules, signature phrases, bilingual rulesBefore every draft
messaging-framework.mdCentral message, per-persona messages, top 10 key numbers, CTA patterns, proof hierarchyBefore every conversion-oriented piece
personas.md2-4 personas with goals, frustrations, expectations, channels, main message per personaBefore any targeted content
brand-platform.mdFull strategic document (mission, vision, positioning, values, architecture)For deep context
style-guide.mdColors, typography, logo usage, components, illustration style, banned visual tropesBefore any visual creation
tokens.jsonPalette, gradient, fonts, type scale, radii, logo clear space in DTCG format: the single machine-readable source. CSS consumers are generated from it (python3 scripts/build-tokens.py)Before any CSS, composition or export. Never type a hex code from memory
anti-ai-writing-style.mdAnti-AI doctrine: writing and formatting rules, banned list (dead AI vocabulary, negative parallelisms, rule of three, em dash)Before every draft, and again at review. One tell is enough to block
exemples-rejetes.mdRejection corpus: one entry per rejected draft (faulty excerpt in a code block, reason, rule with its section, retained fix), fed by step 5 of brand-checkWith the published examples, before drafting. The archives say what the brand does, this file what it refused
divulgation-ia.mdAI disclosure policy: principle, disclosure per content type, wording per channel, généré-par-ia: field of the asset catalogue, what is never generated, edge casesBefore publishing a generated visual, voice or video. Single reference: other files point here
droits.mdRights register: font licences, third-party logo permissions, image rights of people, AI-derived portraits, and the human actions still pendingBefore publishing a client or partner logo, a photo of a person, a portrait. "à confirmer" is an open question, not a permission
stakeholders.mdFounders, team, functional rolesWhen citing a person or assigning an action
assets/Logos, banners, illustrations, photosReuse existing visuals before generating new ones

Universal brand rules (condensed)

Synthesis of voice.md and style-guide.md. If you only have a minute, this is what matters.

Tone

{{BRAND_VOICE_POSITION}}

Vocabulary

  • ✅ Prefer: {{BRAND_VOCABULARY_PREFERRED}}
  • ❌ Avoid / ban: {{BRAND_VOCABULARY_BANNED}}

Typography and colors

  • Primary font: {{BRAND_FONT_PRIMARY}}
  • Primary: {{BRAND_COLOR_PRIMARY}}
  • Accent: {{BRAND_COLOR_ACCENT}}
  • Dark: {{BRAND_COLOR_DARK}}
  • Light: {{BRAND_COLOR_LIGHT}}
  • Signature gradient: {{BRAND_GRADIENT}}
  • Border-radius: {{BRAND_BORDER_RADIUS}}

Visuals

  • ✅ Prefer: {{BRAND_ILLUSTRATION_STYLE}}
  • ❌ Ban: {{BRAND_BANNED_VISUALS}}

Signature phrases and taglines

{{BRAND_TAGLINES}}

Top key numbers

{{BRAND_TOP_NUMBERS}}

Access

Read the files directly — this folder is small enough that file reads are cheap. voice.md is short and often read whole; for a targeted lookup (a number, a banned word), grep the relevant file (messaging-framework.md, voice.md) instead of reading everything.

What the skills never judge from memory

The brand is checkable by script; these checks take precedence over any judgment call.

QuestionFile or command
Exact value of a color, a font, a radius?tokens.json; python3 scripts/build-tokens.py --check
Does this text follow the charter (banned words, dashes, titles, hashtags, placeholders, palette, fonts)?python3 scripts/lint-brand.py <file> (step 0 of brand-check)
Has this topic already been covered, on which channel, when?_templates/inventory.md; python3 scripts/build-inventory.py --check
May we publish this font, logo or photo?droits.md
Must AI involvement be disclosed, and how?divulgation-ia.md
Was this phrasing already rejected?exemples-rejetes.md

Brand consistency filter (5 points)

Before any content ships, the producing role must pass this filter — this is what the brand-check skill enforces. Step 0 runs first: python3 scripts/lint-brand.py <draft> (any error blocks).

  1. Vocabulary — no banned word; preferred vocabulary present where relevant
  2. Tone — aligned with {{BRAND_VOICE_POSITION}}
  3. Proof — every major claim backed by a number from messaging-framework.md or a cited external reference
  4. Audience — target persona identifiable; main message matches their expectation
  5. Visual / format — colors, font, border-radius match style-guide.md

Fail on any point → back to drafting, no publication.

Conflicts and updates

  • If another CLAUDE.md contradicts this folder: 01-brand/ wins. Surface the conflict.
  • If a rule is missing: do not invent. Escalate to {{COMPANY_MAIN_CONTACT}}.
  • If a color or font changes: edit tokens.json, run python3 scripts/build-tokens.py, never patch a CSS file by hand.
  • If a word joins the banned list: add it to voice.md, then to [forbidden-words.charter] of scripts/lint-brand.toml.
  • If a number becomes outdated: update messaging-framework.md and brand-platform.md together.

What this folder does NOT do

  • ❌ Produce content (→ roles 03 through 09)
  • ❌ Manage the editorial calendar (→ 02-strategy/ + {{EDITORIAL_CALENDAR_TOOL}})
  • ❌ Create visuals (→ 06-graphic-design/ + image-generation skill)
  • ❌ Host the skills themselves (→ .claude/skills/) — but every skill must read 01-brand/ before acting.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.