agentleFS
itqanlabitqanlab/agent-toolkit/.agents/skills/higgsfield-studio-brandkit/SKILL.md

higgsfield-studio-brandkit

Skill1 starsChanged 34 days agoMITView on GitHub

Build a complete brand identity with Higgsfield: logo marks, palette, type, mockups, packaging, signage, social graphics and an editable brandbook. Triggers: 'create a brand kit', 'make a visual identity', 'design a logo and brandbook', 'apply this logo to branded assets', 'make packaging or signage', 'extend our existing branding'. Part of Higgsfield Studio: every image is priced first, never paid for twice, and written to a ledger. Needs Node 18+, Python 3 and the higgsfield command line tool.

The file

Show its rules and text

We split this skill into the rules it gives an agent

23 rules. About 0 are common, 17 are distinctive, and 6 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈17

    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:

    Put --max N (a little above the price), --tag and --why "..." on every paid job.

    line 37
  2. Very specific:

    Ask before spending more than about 10 credits, or 10% of what is left.

    line 38
  3. Distinctive:

    Read prerequisites. Check tools before the stage that needs them.

    line 66
  4. Very specific:

    If higgsfield is missing, install it only after permission:

    line 67
  5. Very specific:

    If higgsfield account status fails with an authentication or workspace error, ask the user to run higgsfield auth login or select a …

    line 73
  6. Distinctive:

    Ask one compact set of only unresolved blocking questions.

    line 103
  7. Distinctive:

    After each palette, logo, typography, or downstream review, stop and wait for ordinary user feedback.

    line 104
  8. Distinctive:

    Never infer approval from silence, successful generation, or your own preference.

    line 105
  9. Distinctive:

    Run intake and asset analysis: Read intake, asset analysis, state routing, and exact state payloads.

    line 122
  10. Distinctive:

    Create the Brand Lock: Read Brand Lock. Record exact spelling, official assets, colors, fonts, layout/shape rules, requested outputs, and …

    line 123
  11. Distinctive:

    Build missing foundation slots: Read Design Brain, concept boards, inline reviews, and only the needed palette,

    line 131
  12. Distinctive:

    QA and approval: Read QA and iteration. Repair only the failing output.

    line 142
  13. Distinctive:

    Read logo prompt enhancer.

    line 152
  14. Distinctive:

    Use the returned SVG URLs directly for review.

    line 164
  15. Distinctive:

    Use Recraft V4.1 vector mode only for new logo marks.

    line 185
  16. Distinctive:

    Use Seedream as the primary photoreal mockup generator.

    line 186
  17. Very specific:

    Use local deterministic SVG/PPTX/HTML construction for exact copy and editable layouts.

    line 187
  18. Distinctive:

    Do not promise native Figma, Canva, PSD, AI, or EPS files.

    line 188
  19. Very specific:

    Create JSON input files under "$BRANDKIT_WORKDIR".

    line 192
  20. Distinctive:

    If preview or logo export fails twice, report the concrete error.

    line 214
  21. Distinctive:

    If the Brandbook template, font, or conversion contract fails, stop immediately.

    line 215
  22. Distinctive:

    If exact typography or official-logo fidelity cannot be preserved, disclose the limitation instead of claiming completion.

    line 216
  23. Distinctive:

    Never expose raw auth tokens or credentials in files, logs, or chat.

    line 217

≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill 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
SKILL.md

Higgsfield Brand Kit

A brand system built in stages, with approvals saved in the project, so only what depends on a change is made again. Logos are drawn with recraft_v4_1 (quoted at 1.25 credits on 2026-10-12), mockups and scenes with other image models, and the brandbook is assembled locally by the Python tools in scripts/, which cost nothing.

What it costs

A full kit is many images. Before starting, list the deliverables the user wants, price one image of each kind with --dry, and give the user a total. Make logo candidates one at a time and let the user pick before any mockup is made.

Before anything else

sh ../higgsfield-studio/scripts/setup-deps.sh --check
node ../higgsfield-studio/scripts/hf.mjs check

The first checks Node and the higgsfield tool, and sh ../higgsfield-studio/scripts/setup-deps.sh installs what is missing after asking. The second checks the sign-in. If Higgsfield is not signed in, the user signs in once with higgsfield auth login, which opens the browser. Never print the sign-in token or open the CLI's credential file.

Spending credits

This skill is part of the Higgsfield Studio plugin. Every paid command runs through the main skill's guard, node ../higgsfield-studio/scripts/hf.mjs. It asks Higgsfield for the price before running, refuses above --max or below the reserve, never pays twice for the same request, and writes what each job really cost to .higgsfield/ledger.jsonl in the project. The full rules are in ../higgsfield-studio/SKILL.md. The short form:

  • Price first with --dry, using every parameter the real job will use. Tell the user the total for anything over about 5 credits.
  • Put --max N (a little above the price), --tag and --why "..." on every paid job.
  • Ask before spending more than about 10 credits, or 10% of what is left. node ../higgsfield-studio/scripts/hf.mjs account shows the balance.
  • Anything outward or not undoable goes through act and needs --yes. Say what will happen and wait for the user's yes before adding it.
  • Report in plain words, in the user's language: where the file is, what it cost, what is left. No JSON or job ids unless asked.

Free commands, such as lists, model get, presets and uploads, can run as node ../higgsfield-studio/scripts/hf.mjs run <command> or as higgsfield <command> directly.

Updates

This skill is versioned with its plugin. Its version is metadata.version in the header of this file. The plugin's history, for every skill in it, is in ../higgsfield-studio/CHANGELOG.md. To check for a newer version, open the address on its Latest: line and compare the top version with this one. Read every entry in between and tell the user what changed before anything is updated.

Higgsfield's guide

Everything below is Higgsfield's official guide for this task, with its paid and outward commands already pointed at node ../higgsfield-studio/scripts/hf.mjs. The rules above win where the two differ.

Higgsfield Brandkit

Build a coherent identity and its requested applications. Treat supplied brand facts and official assets as fixed constraints.

Bootstrap

  1. Resolve SKILL_ROOT to this skill's installed directory and create a durable project directory:

    BRANDKIT_WORKDIR="${PWD}/brandkit"
    BRANDKIT_STATE="${BRANDKIT_WORKDIR}/state.json"
    mkdir -p "${BRANDKIT_WORKDIR}"
  2. Read prerequisites. Check tools before the stage that needs them. Never install system packages without the user's permission.
  3. If higgsfield is missing, install it only after permission:

    sh ../higgsfield-studio/scripts/setup-deps.sh
  4. If higgsfield account status fails with an authentication or workspace error, ask the user to run higgsfield auth login or select a workspace, then wait.
  5. Inspect live model contracts before paid generation:

    higgsfield model get recraft_v4_1 --json
    higgsfield model get seedream_v5_pro --json
    higgsfield model get gpt_image_2 --json

CLI mapping

OperationCommand
Discover a modelhiggsfield model get <model> --json
Generate and pollnode ../higgsfield-studio/scripts/hf.mjs create <model> ... --wait --json
Resume a jobhiggsfield generate wait <job_id> --json
Upload a local assethiggsfield upload create <path> --json
Import website metadatahiggsfield marketing-studio brand-kits fetch --url <url> --wait --json
Read/write approval statepython3 "$SKILL_ROOT/scripts/brandkit.py" state ...
Render review boardspython3 "$SKILL_ROOT/scripts/brandkit.py" preview ...
Inspect selected logopython3 "$SKILL_ROOT/scripts/brandkit.py" logo-inspect ...
Export logo filespython3 "$SKILL_ROOT/scripts/brandkit.py" logo-export ...
Build a Brandbookpython3 "$SKILL_ROOT/scripts/brandkit.py" brandbook-build ...

Local image paths passed with --image are auto-uploaded. Keep HTML, SVG, PPTX, and PDF deliverables as local project files unless the user explicitly needs a hosted copy.

User-facing behavior

  • Match the user's language. Keep Design Brain reasoning, prompts, state mechanics, scripts, model lookup, and QA internals private.
  • Send at most one short status sentence per visible generation batch, then stay quiet until the result is ready.
  • Ask one compact set of only unresolved blocking questions. Never repeat facts or force a complete identity questionnaire for a partial task.
  • After each palette, logo, typography, or downstream review, stop and wait for ordinary user feedback.
  • Never infer approval from silence, successful generation, or your own preference.
  • Preserve exact user copy. Never invent positioning, values, claims, ingredients, prices, certifications, statistics, or regulatory content.

Core workflow

  1. Classify the request.
    • apply-existing: use supplied official assets without redesigning them.
    • extend-partial: create only missing slots required by the requested output.
    • create-identity: create a new logo or identity only when explicitly requested.
  2. Read state. Run:

    python3 "$SKILL_ROOT/scripts/brandkit.py" state \
      --state-file "$BRANDKIT_STATE" --action get_status

    Local state is durable. Never paste, hand-edit, or recreate approvals when the state file exists.

  3. Run intake and asset analysis. Read intake, asset analysis, state routing, and exact state payloads. Lock every user-declared official logo, palette, and typography slot immediately.
  4. Create the Brand Lock. Read Brand Lock. Record exact spelling, official assets, colors, fonts, layout/shape rules, requested outputs, and forbidden treatments.
  5. Require only the slots the output uses.
    • logo-only → palette + logo for a new mark; official logo alone for an existing mark
    • palette-only → palette
    • typography-only → typography
    • copy-free mockup/merch → logo; add palette only when color/application requires it
    • text-bearing social/packaging/poster/signage → logo + palette + typography
    • Brandbook/deck → logo + palette + typography
  6. Build missing foundation slots. Read Design Brain, concept boards, inline reviews, and only the needed palette, logo, or typography module.
  7. Continue the original request as soon as its required slots are approved. Never ask the user to choose scope again.
  8. Load only the requested production module:
    • mockups
    • social graphics
    • posters/banners
    • packaging
    • signage
    • merchandise
    • presentation decks
    • Brandbooks
  9. QA and approval. Read QA and iteration. Repair only the failing output. Save a downstream element only after explicit approval with its exact foundation dependencies.

New identity sequence

1. Palette

Render 2–3 exact palette options as deterministic HTML using preview payloads. Show PNG screenshots plus editable HTML files and wait. Persist the selected palette with approve_palette before logo generation.

2. SVG logo marks

Read logo prompt enhancer. Produce exactly three distinct symbol-only mechanisms and one Recraft prompt for each. Write each long prompt to a file and submit separately:

node ../higgsfield-studio/scripts/hf.mjs create recraft_v4_1 \
  --model_type vector \
  --colors @"${BRANDKIT_WORKDIR}/logo-colors.json" \
  --background_color '#F7F7F5' \
  --aspect_ratio 1:1 \
  --resolution 2k \
  --prompt-file "${BRANDKIT_WORKDIR}/logo-candidate-1.txt"

Use the returned SVG URLs directly for review. After selection, inspect the exact SVG without altering it:

python3 "$SKILL_ROOT/scripts/brandkit.py" logo-inspect \
  --source "<selected Recraft SVG URL or absolute local path>"

Persist the exact job ID, SVG URL, name, palette revision, and returned canonical geometry fingerprint with approve_logo.

3. Typography

Propose 2–3 unique display/body pairs using supplied fonts or verified Google Fonts. Render the real brand name and sample copy through the preview script. Persist only the selected pair with approve_typography.

Interactive flows always stop for palette, logo, and typography selections. Explicit no-question mode may choose and persist a palette, but it still shows all three SVG logo candidates and stops for the user's logo selection; exact brand marks are never self-approved.

Consistency invariants

  • Reuse the same approved logo source everywhere. Never redraw an official or selected SVG when deterministic placement/export is possible.
  • A generated logo depends on the palette revision used to create it. Changing that palette invalidates the generated logo and its dependents; changing typography does not invalidate the symbol mark.
  • Changing a foundation slot invalidates only downstream elements that list that slot in required_slots.
  • Copy the same Brand Lock values into every related generation prompt: exact hex, font roles, shape language, placement, clear space, composition, and forbidden treatments.
  • Use Recraft V4.1 vector mode only for new logo marks.
  • Use Seedream as the primary photoreal mockup generator. Use GPT Image 2 only for the controlled stage that adds readable text or exact graphic details.
  • Use local deterministic SVG/PPTX/HTML construction for exact copy and editable layouts. Do not ask an image model to fake editable files.
  • Do not promise native Figma, Canva, PSD, AI, or EPS files.

Deterministic scripts

Create JSON input files under "$BRANDKIT_WORKDIR"; never interpolate user text directly into shell arguments.

python3 "$SKILL_ROOT/scripts/brandkit.py" preview \
  --input "$BRANDKIT_WORKDIR/reviews.json" \
  --output-dir "$BRANDKIT_WORKDIR/reviews"

python3 "$SKILL_ROOT/scripts/brandkit.py" logo-export \
  --input "$BRANDKIT_WORKDIR/logo-export.json" \
  --output-dir "$BRANDKIT_WORKDIR/logo"

python3 "$SKILL_ROOT/scripts/brandkit.py" brandbook-build \
  --state-file "$BRANDKIT_STATE" \
  --input "$BRANDKIT_WORKDIR/brandbook.json" \
  --output-dir "$BRANDKIT_WORKDIR/brandbook"

For logo export, load logo export payloads. For Brandbooks, use the bundled builder only; never substitute an improvised PowerPoint or PDF generator after a deterministic contract failure.

Failure policy

  • Retry a failed Recraft or image-generation request once with the same locked concept and corrected contract. Stop after the second equivalent failure.
  • If preview or logo export fails twice, report the concrete error; never replace it with ad-hoc SVG rewriting.
  • If the Brandbook template, font, or conversion contract fails, stop immediately. Do not produce a visually different fallback and call it canonical.
  • If exact typography or official-logo fidelity cannot be preserved, disclose the limitation instead of claiming completion.
  • Never expose raw auth tokens or credentials in files, logs, or chat.

Delivery

For Brandbooks, follow the strict response contract in brandbook: PPTX link/path, PDF link/path, and font-install warning only.

For other outputs return:

  1. The requested visual files and previews.
  2. A compact Brand Lock summary.
  3. Editable versus flattened format labels.
  4. Required-font/import limitations.
  5. Stable variant names for targeted revisions.

Reference index

  • Prerequisites — stage-specific local dependencies and install commands.
  • Intake — minimal questions and input routing.
  • Asset analysis — official/reference classification and measurement.
  • State routing and state payloads — persistent approvals.
  • Brand Lock — canonical visual constraints.
  • Design Brain — private art direction.
  • Concept boards, preview payloads, and inline reviews — selection stages.
  • Logo, logo prompt enhancer, and logo export payloads — SVG generation and deterministic variants.
  • Palette and typography — foundation slots.
  • Mockups, social graphics, posters/banners, packaging, signage, and merchandise — applications.
  • Presentation decks and Brandbooks — editable documents.
  • QA and iteration — preflight, repair, approval, and delivery manifest.

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.