brand-skill-builder
Turn a brand book into AI skills that make on-brand files and audit them, on Claude and ChatGPT.
Safety scan
A static text scan, not a guarantee. Select the badge to see what was scanned and every finding.
Badge for your README
Markdown
[](/listings/yuckier--brand-skill-builder)HTML
<a href="/listings/yuckier--brand-skill-builder"><img src="/api/listings/yuckier--brand-skill-builder/badge.svg" alt="findsafeskills safety rating" /></a>Install
Install instructions from the README
As written by the repo's author. findsafeskills has not verified these steps.
**Claude:** zip the `skills/brand-skill-builder` folder and upload it as a skill, or copy the folder into your skills directory (for Claude Code, `~/.claude/skills/`).
**ChatGPT and other agents:** the skill ships `agents/openai.yaml`; load the folder as a skill wherever your platform supports the Agent Skills format.
The skill works best where the model has a persistent file system (Claude Code, Codex, a local agent), because it writes a whole library of files.
What was scanned
The files the safety scan read, as written by the repository's author and shown as plain text. Lines with a finding are marked.
README.md · README · 69 lines
- # Brand Skill Builder
- A skill that teaches an AI model to build **brand skills**: a small library of skills that produce on-brand work for one
- company, from that company's brand book.
- It comes from a production library of brand skills (voice, PDF, Word, PowerPoint, social images, HTML, motion and a brand
- audit) that runs on both Claude and ChatGPT. This repository holds the method, the templates and the tools from that
- library, with no brand's rules in it.
- ## What you get
- - **A workflow** (`SKILL.md`) for three jobs: start a new library from a brand book, add a skill to a library, or review an
- existing brand skill against the pattern.
- - **References** covering the architecture, how to turn a brand book into structured foundations, the anatomy of a format
- skill, the voice and audit skills, QA and testing, packaging, and lessons from the build.
- - **Templates** for the library (`AGENTS.md`, `DEV-LOG.md`), the brand foundations (palette, typography, structural system,
- accessibility, voice), each skill type (format, voice, audit) and a format's spec and map.
- - **Scripts** (Python 3.9 or newer, standard library only):
- - `scaffold.py`: creates a library or a new skill from the templates
- - `verify_context.py`: the preflight every skill runs (digests, brand era, review date, platform folders)
- - `snapshot.py`: refreshes a skill's bundled context and manifest from the shared source
- - `package_skill.py`: packages a skill into an upload zip and checks it (strict frontmatter, description length,
- digests, no nested zips, clean extraction)
- ## The approach in six lines
- 1. **The brand book becomes data.** Palette, typography, structural system, accessibility and voice, with provenance, and
- every gap recorded instead of guessed.
- 2. **One skill, one job.** Voice, each output format and the audit are separate skills that never call each other at runtime.
- 3. **Snapshot-first.** Each skill bundles the brand context it needs, with SHA-256 digests, so it runs offline and can prove
- which version of the brand it used.
- 4. **The model plans, a script builds, a script checks.** The model writes a JSON spec; a builder reads brand tokens at build
- time; a QA gate checks the finished file against the brand's ship check.
- 5. **Warn, never refuse.** Skills repair and report, deliver failing files labelled, and stop only when something is
- physically impossible.
- 6. **House decisions are written down.** Every choice the brand book did not make is recorded with its scope, its status
- and who decided.
- ## Install
- **Claude:** zip the `skills/brand-skill-builder` folder and upload it as a skill, or copy the folder into your skills
- directory (for Claude Code, `~/.claude/skills/`).
- **ChatGPT and other agents:** the skill ships `agents/openai.yaml`; load the folder as a skill wherever your platform
- supports the Agent Skills format.
- The skill works best where the model has a persistent file system (Claude Code, Codex, a local agent), because it writes a
- whole library of files.
- ## Quick start
- ```bash
- python3 skills/brand-skill-builder/scripts/scaffold.py library ./acme-skills --brand "Acme" --era 2027-brand
- python3 skills/brand-skill-builder/scripts/scaffold.py skill ./acme-skills acme-pdf --kind format --format PDF
- python3 acme-skills/tools/snapshot/snapshot.py acme-skills/tools/snapshot/acme-pdf.json
- python3 acme-skills/skills/acme-pdf/scripts/verify_context.py
- python3 acme-skills/skills/acme-pdf/tests/test_platforms.py
- ```
- Then ask your model: "Use brand-skill-builder to fill the Acme foundations from the brand book in `acme-skills/reference/`."
- ## What it does not include
- No brand's colours, fonts, voice or rules, and no format builders: a builder is specific to a brand's visual system, so each
- library writes its own, following `references/format-skills.md`.
- ## Licence
- MIT. See [LICENSE](LICENSE).
Ratings and comments
Sign in to rate this skill.
Sign inFree to use. If it’s helped, pass it on.