Skills
Build and govern a brand system with ak:brand
Turn approved brand inputs into reviewable guidelines, design tokens, messaging rules, and asset checks without treating generated choices as evidence.
Use ak:brand to define or update a brand's voice, messaging, visual identity,
and asset rules. The Skill can guide brand work, and its documented update
route can synchronize approved guideline values into local design-token files.
Choose this Skill for a governed brand artifact
Use ak:brand when
- You need a voice chart, message hierarchy, positioning statement, visual rules, or brand-guideline draft.
- You want to review an asset against approved naming, format, size, palette, accessibility, licensing, and messaging criteria.
- You have approved color and mood inputs and want to update the local brand authority and derived tokens together.
Choose another workflow when
- Audience, market, or competitor claims are still unknown. Start with ak:marketing-research, ak:competitor, or ak:persona, then distinguish evidence from assumptions.
- You need a campaign plan rather than a brand system. Use ak:marketing-planning.
- You only want to publish, generate media, or change a live design system. Those are separate actions with their own approval and provider effects.
Prepare the inputs and authority
Provide the business and product facts, target audience evidence, existing
assets, approved terms, legal constraints, and the person who can approve brand
changes. By default, the bundled helpers treat docs/brand-guidelines.md as the
local source of truth. Confirm that this is the correct project authority before
allowing writes.
Separate supplied facts from proposed choices. A color, voice trait, proof point, or differentiator generated by the workflow is a draft until an owner approves it and any factual claim is supported.
Invoke the documented update route
| Runtime | Invocation | Availability boundary |
|---|---|---|
| Claude Code | /ak:brand update ... | Native delivery is the default; explicit plugin delivery is also supported. |
| Cursor | /ak:brand update ... | Slash invocation is user-verified; this does not establish broader runtime parity. |
| Codex | $ak:brand update ... | Uses native Skill discovery; supporting Hooks and scripts can differ from Claude Code. |
/ak:brand update "Update the draft theme to the approved Ocean Blue, Golden Amber, and Emerald values. Keep unverified messaging as assumptions and show every file before final approval"/ak:brand update "Update the draft theme to the approved Ocean Blue, Golden Amber, and Emerald values. Keep unverified messaging as assumptions and show every file before final approval"$ak:brand update "Update the draft theme to the approved Ocean Blue, Golden Amber, and Emerald values. Keep unverified messaging as assumptions and show every file before final approval"The argument hint also names review and create, but this package includes a
dedicated route only for update. State review or creation work as a natural-
language outcome unless your installed package supplies an additional route.
Understand the stages
- Establish authority. Confirm the current brand guideline, audience and product evidence, approval owner, and permitted file scope.
- Draft the system. Develop voice, tone, message architecture, visual identity, asset organization, and approval criteria as required.
- Update the guideline. The
updateroute collects theme, primary, secondary, accent, and mood inputs and editsdocs/brand-guidelines.md. - Synchronize tokens. The helper updates
assets/design-tokens.jsonand attempts to regenerateassets/design-tokens.css. A dry-run exists in the helper, but the routed workflow describes the writing path. - Verify the result. Extract the resulting context, inspect the diff, test contrast and asset use, and obtain human approval before use or publication.
Optional helpers can extract brand context, validate an asset's filename, format, size, and manifest registration, or prepare color-comparison steps. Color extraction from an image may require ImageMagick or an external multimodal service; that access is separate from local brand editing.
Keep approval and side effects explicit
Brand generation is not brand approval
The workflow can propose and write brand artifacts. It cannot establish that a claim is true, an asset is licensed, a mark is legally clear, a palette is accessible in every context, or a generated identity is approved.
The update workflow mutates local files. Review all three expected paths and
preserve unrelated changes. Do not treat token generation as permission to
publish, deploy, replace production assets, call an image provider, or incur
cost. Give those permissions separately.
Verify outputs and evidence
A complete run should provide:
- The brand-guideline diff and the approved inputs behind each changed value.
- The derived JSON and CSS token diffs, including any CSS-generation warning.
- Voice, messaging, and visual rules with proof points linked to their sources or labeled as assumptions.
- Asset-check findings separated into errors, warnings, and suggestions.
- Remaining accessibility, legal, licensing, stakeholder, and publication decisions.
Troubleshoot and know the limits
| Symptom | Safe next step |
|---|---|
| The Skill route is not recognized | Confirm target and scope, restart the runtime, then follow Runtime cannot find a Skill or Agent. |
docs/brand-guidelines.md is missing | Create or identify the approved authority before running sync; do not let placeholder values become brand decisions. |
| Token sync reports incomplete colors or CSS failure | Stop, inspect guideline formatting and the JSON diff, then verify the design-system generator exists before retrying. |
| Asset validation passes | Still review color use, accessibility, rights, claims, and context; the helper checks only a bounded set of file rules. |
| Image color analysis is requested | Confirm the image may be sent to the selected tool and approve any network, credential, or provider cost first. |
The bundled templates contain illustrative colors, fonts, ratios, and wording; they are not facts about your brand. Helper parsing depends on expected Markdown sections, and approximate RGB distance is not a complete brand or accessibility test. Continue with the Marketing Kit overview or Marketing Kit lifecycle for installation ownership and recovery.
Build evidence-labeled personas with ak:persona
Create, analyze, update, or list customer personas while keeping observed behavior, supplied facts, hypotheses, and sensitive data clearly separated.
Generate testable marketing options with ak:marketing-ideas
Select a small set of context-fit ideas from the bundled SaaS catalog, expose rationale and resource needs, and turn assumptions into reviewable validation steps.