brand-compliance（品牌合规性）

Brand Compliance Skill

When to Invoke

Invoke this skill when:

• Reviewing marketing copy, blog posts, product descriptions, or social content
• Auditing design assets, landing pages, or campaign materials
• Checking brand guideline compliance before content goes live
• Evaluating cross-channel consistency across touchpoints

Skill({ skill: 'brand-compliance' });

Execution Process

Step 1: Style Guide Validation

Check written content against the brand's style guide:

Grammar and Language:

• Preferred terminology vs. prohibited terms
• Capitalization rules (product names, features, company name)
• Punctuation style (Oxford comma, em-dash usage, etc.)
• Abbreviation and acronym policy

Formatting:

• Heading hierarchy and casing
• List punctuation and structure
• Number formatting (when to spell out vs. use digits)
• Date and time format conventions

Output: List each violation with exact text, rule violated, severity, and suggested fix.

Step 2: Tone of Voice Checking

Score content against the brand voice profile across key dimensions:

Voice Dimensions (score 1–5 per dimension):

• Formality: 1 (casual/conversational) — 5 (formal/corporate)
• Warmth: 1 (distant/clinical) — 5 (approachable/human)
• Authority: 1 (tentative) — 5 (confident/expert)
• Energy: 1 (calm/measured) — 5 (bold/energetic)

Tone Analysis:

1. Extract 3–5 representative passages from the content
2. Score each passage on the four dimensions
3. Compare against the target voice profile
4. Flag passages that deviate >1 point from target
5. Provide 2–3 rewrite suggestions for flagged passages

Default voice profile (use when brand profile not provided):

• Formality: 3, Warmth: 4, Authority: 4, Energy: 3

Step 3: Visual Identity Audit

When reviewing design assets or code with visual specifications:

Logo Usage:

• Correct logo variant (primary, secondary, icon-only)
• Minimum size requirements
• Clear space / exclusion zone compliance
• Prohibited alterations (rotation, distortion, color changes)

Color Palette:

• Only approved palette colors used
• Correct hex/RGB values (flag approximations)
• Accessibility contrast ratios (WCAG AA minimum)

Typography:

• Approved font families only
• Correct weight/style combinations
• Type scale hierarchy compliance

Imagery and Icons:

• Photography style consistency
• Icon language consistency (line weight, style, corner radius)

Step 4: Brand Asset Management Review

Verify asset usage compliance:

Asset Categories:

• APPROVED: Current, compliant versions
• DEPRECATED: Old versions — flag for replacement
• RESTRICTED: Context-limited assets — verify usage context

Checks:

• Assets sourced from approved DAM or asset library
• Version currency (not using outdated logo versions)
• Usage rights verified for photography/illustrations

Step 5: Cross-Channel Coherence

Evaluate consistency across channels when multiple touchpoints are provided:

Core Brand Elements (must be consistent across all channels):

• Brand name formatting
• Key taglines and value propositions
• Color palette adherence
• Typography choices

Channel-Specific Adaptations (acceptable variations):

• Image aspect ratios
• Copy length
• CTA phrasing for platform norms

Coherence Score: Rate 1–5 per channel pair reviewed.

Step 6: Generate Compliance Report

Produce a structured report:

# Brand Compliance Report

**Content/Asset:** [name]
**Date:** [date]
**Audited By:** brand-compliance skill

## Summary

- Total Findings: N
- Errors (Blocking): N
- Warnings (Informational): N
- Passed: N

## Findings

| #   | Area   | Finding                              | Severity | Suggested Fix      |
| --- | ------ | ------------------------------------ | -------- | ------------------ |
| 1   | Style  | [exact text] uses "..." (prohibited) | ERROR    | Replace with "..." |
| 2   | Tone   | Formality score 5 vs. target 3       | WARNING  | Rephrase: "..."    |
| 3   | Visual | Logo minimum size violated           | ERROR    | Use 40px minimum   |

## Tone Score

| Dimension | Target | Actual | Delta | Status |
| --------- | ------ | ------ | ----- | ------ |
| Formality | 3      | 4      | +1    | WARN   |
| Warmth    | 4      | 4      | 0     | PASS   |
| Authority | 4      | 3      | -1    | WARN   |
| Energy    | 3      | 3      | 0     | PASS   |

## Overall Compliance Score

[N]% compliant (errors weighted 3x, warnings weighted 1x)

## Next Steps

1. [Action for error #1]
2. [Action for error #2]

Example 1: Copy Audit

User: "Review this product description for brand compliance"
Agent: [Reads content → runs Steps 1–2 → produces findings table with severity]
Output: 3 errors (prohibited terms, wrong capitalization), 2 warnings (tone slightly too formal), 92% compliance score

Example 2: Visual Asset Audit

User: "Check this landing page design for brand compliance"
Agent: [Reads design specs/CSS → runs Steps 3–5 → produces visual findings]
Output: 1 error (off-palette color #E5E5E5 used), 3 warnings (typography weight inconsistency), 88% compliance

Example 3: Cross-Channel Review

User: "Compare our Instagram and web copy for brand coherence"
Agent: [Reads both → runs Step 5 → coherence score]
Output: 85% coherent, tagline wording differs across channels (warning), CTA variations acceptable

Enforcement Hooks

Input validated against schemas/input.schema.json before execution. Output contract defined in schemas/output.schema.json. Pre-execution hook: hooks/pre-execute.cjs Post-execution hook: hooks/post-execute.cjs (emits observability event)

Iron Laws

1. ALWAYS validate before publish, not after — catching violations at creation time reduces rework by 65%; post-publish fixes compound reputation cost.
2. ALWAYS score tone against an explicit voice profile — subjective tone judgments without a baseline are inconsistent and contestable; require a target profile as input.
3. ALWAYS separate error (blocking) from warning (informational) — blocking on warnings creates friction that causes teams to disable compliance checks entirely.
4. ALWAYS provide specific remediation, not just flags — reports with 2–3 concrete rewrite suggestions have near-100% resolution rates vs. 30% for "tone is off" messages.
5. ALWAYS use design tokens as ground truth for visual checks — check against tokens.json or approved color/type files when available, not just memory.

Anti-Patterns

Memory Protocol (MANDATORY)

Before starting: Read .claude/context/memory/learnings.md

After completing:

• New brand pattern discovered → .claude/context/memory/learnings.md
• Issue with brand guideline gaps → .claude/context/memory/issues.md
• Brand voice profile decision → .claude/context/memory/decisions.md

ASSUME INTERRUPTION: If it's not in memory, it didn't happen.