✅ /strategy/          Brand bible (polished, timeless)
✅ /research/          Research domains (temporal, date-stamped)
✅ /docs/              Architecture documentation
✅ .claude/            Architecture configuration
   ✅ /output-styles/  Primary agent behavior
   ✅ /agents/         Sub-agent definitions
   ✅ /skills/         Capabilities (progressive disclosure)
   ✅ /commands/       Workflow triggers
✅ .mcp.json          Tool integrations
✅ CLAUDE.md          Project instructions

Current: Marketing architects write skill files manually
Desired: Meta command generates structure, ensures compliance

Current: Architectural compliance is manual review
Desired: Build-time checks, CI/CD validation

Current: Plan/implement are conceptual patterns
Desired: Actual meta commands that scaffold and track

Current: All work goes directly to execution
Desired: Optional human review gates for quality control

Creates PLAN.md with:
  - Approach outline
  - Resources needed (agents, skills, files, tools)
  - Estimated effort
  - Deliverables

Waits for Marketing Architect approval

Prerequisites: Approved PLAN.md exists

Creates TODO.md
Executes plan step-by-step
Tracks progress
Flags blockers
Delivers results

Scaffolds research domain:
  /research/{domain}/
    ├── RESEARCH.md (template)
    ├── /data/
    ├── /execution/
    └── /exports/

Guided setup for domain scope

✅ Checks for contradictions between parent and child
✅ Validates frontmatter metadata
✅ Ensures footnote format is correct
✅ Verifies referenced research files exist

✅ Detects circular dependencies
✅ Ensures one-way dependencies
✅ Validates skills reference existing tools
✅ Checks agents reference existing skills

✅ Warns if SKILL.md > 500 lines
✅ Checks for entry points (RESEARCH.md, STRATEGY.md)
✅ Validates file hierarchy depth (max 3 levels)

Marketing Architect: "create skill: Analyzing Social Media Sentiment"

System prompts:
  "Describe the methodology for this skill..."

Marketing Architect explains approach in conversation

Meta command generates:
  .claude/skills/analyzing-social-sentiment/
    ├── SKILL.md (< 500 lines, overview)
    ├── platform-analysis.md (Twitter, LinkedIn, etc.)
    ├── sentiment-scoring.md (methodology)
    └── reporting-framework.md (outputs)

Validates:
  - Progressive disclosure compliance
  - One-way dependencies
  - File structure conventions

Result: Skill ready to use, architecturally compliant

Marketing Architect: "create agent: Campaign Strategist"

System prompts:
  "What should this agent be responsible for?"
  "Which existing skills should it have access to?"
  "What tone/communication style?"

Meta command generates:
  .claude/agents/campaign-strategist.md:
    Role: Campaign Strategist
    Purpose: [from conversation]
    Skills: [selected during conversation]
    Tone: [defined during conversation]

Validates:
  - Referenced skills exist
  - No circular dependencies
  - Follows agent definition template

Result: Agent ready to use

Creates review workflow:
  1. Work completed (content, research, campaign)
  2. System generates /review/{date}-{topic}.md with:
     - What's being reviewed
     - Checklist of quality criteria
     - Approval/feedback section
  3. Marketing Architect reviews
  4. Approves OR provides feedback
  5. If feedback: agent iterates
  6. Once approved: marked as final

Marketing Architect: "review: Q1 campaign messaging"

System creates:
  /review/2026-01-15-q1-campaign-messaging.md

Contains:
  - All campaign assets
  - Brand consistency checklist
  - Research backing verification
  - Approval section

Marketing Architect reviews, approves or requests changes

As marketing architects edit strategy files:
  → Check for contradictions in real-time
  → Warn about missing footnotes
  → Flag potential circular dependencies

Before committing to Git:
  → Run full validation suite
  → Block commits with errors
  → Allow commits with warnings (but flag them)

On pull requests:
  → Validate all changed files
  → Show validation results in PR
  → Require passing validation to merge

Generate reports:
  → % of strategy with research backing
  → Orphaned files (no references)
  → Progressive disclosure compliance
  → Dependency graph visualization

Use case: Post campaign updates, share research findings
Skill: "Distributing Content to Channels"
Integration: Post to specific channels, threads

Use case: Reference brand assets, design systems
Skill: "Accessing Brand Visual Assets"
Integration: Pull colors, fonts, logos into context

Use case: Performance-based content iteration
Skill: "Analyzing Content Performance"
Integration: Pull metrics, identify top performers

Use case: Customer language analysis
Skill: "Analyzing Customer Communication"
Integration: Pull support tickets, emails for research

Marketing architects publish skills:
  - "Analyzing Podcast Transcripts"
  - "Creating Video Scripts"
  - "Planning Webinar Sequences"

Other marketing architects can:
  - Browse available skills
  - Adapt for their use
  - Contribute improvements

Share agent definitions:
  - "SEO Specialist"
  - "Email Marketer"
  - "Social Media Manager"

Others adapt for their brand

Share workflow commands:
  - "/product-launch/sequence"
  - "/content-calendar/monthly"
  - "/research/customer-interviews"

Others customize for their processes

GitHub-based initially:
  - Agents/skills/commands in public repos
  - Marketing architects fork and adapt
  - PRs for improvements

Future: Built-in sharing within framework

- Which agents used most frequently
- Which skills are most valuable
- Which commands are most common
- Token usage patterns

- Which brand pillars used most
- Which research references cited most
- Content type distribution
- Quality metrics (human ratings)

- Temporal comparison frequency
- Research domain growth
- Strategy update cadence
- Audit trail strength (% backed by research)

- Time saved vs. manual workflows
- Consistency improvements
- Research utilization rates
- Content output volume

If models support 1M+ tokens:
  - Load more strategy context
  - Analyze entire research domains
  - Generate longer-form content
  - Deeper comparative analysis

If vision/audio/video support improves:
  - Analyze brand visual assets (Figma, images)
  - Transcribe and analyze video content
  - Generate image prompts from brand guidelines
  - Create multimedia campaign assets

Fine-tune models on:
  - Specific brand voice
  - Industry-specific language
  - Company-specific research patterns

Source: Marketing architects using the system
Process:
  - Identify repeated manual workflows
  - Assess if meta command would help
  - Build if broadly useful

Source: Limitations discovered in practice
Process:
  - Identify friction points
  - Design pattern to resolve
  - Implement systematically

Source: Barriers to adoption/usage
Process:
  - Understand barrier
  - Build tool/command to lower it
  - Measure adoption improvement

Source: Marketing architects build & share
Process:
  - Observe patterns in shared components
  - Consider systematizing into core
  - Implement if high impact

✅ New meta commands added (opt-in)
✅ New validation checks (warnings, not errors initially)
✅ New integrations (optional)
✅ New documentation (supplemental)

❌ Changing core architecture (one-way dependencies)
❌ Removing existing capabilities
❌ Forcing new workflows
❌ Breaking existing files

1. Announce early (3+ months notice)
2. Provide migration tooling
3. Support old + new for transition period
4. Document migration path
5. Increment major version (v1 → v2)

1. Use the system
2. Document pain points / friction
3. Share what workflows you wish were systematized
4. Request features in discussions/issues

1. Create skills/agents/commands
2. Share with community
3. Contribute to documentation
4. Suggest meta commands for common patterns

1. Build meta commands from user feedback
2. Improve validation tooling
3. Enhance documentation
4. Ensure architectural integrity
5. Support marketing architects