Skip to main content

​
Layer 4: Skill Layer (Capabilities)

​
What This Layer Is

Skills are reusable workflows and capabilities that agents use to get work done. They’re the “how” of your marketing system. Think of skills like:
  • Standard operating procedures (how we conduct research)
  • Methodologies (how we analyze data)
  • Frameworks (how we structure content)
  • Best practices (how we ensure quality)
Skills are not agents—they’re capabilities that agents (both Operations Manager and sub-agents) can invoke.

​
The Key Innovation: Progressive Disclosure

​
The Problem Skills Solve

Token limits are finite. You can’t load all context into every agent conversation. But you also can’t have agents work without guidance—that creates generic AI slop. The solution: Progressive disclosure.

​
What Is Progressive Disclosure?

Progressive disclosure means: Load only the context that’s relevant to the current task, at the moment it’s needed. Instead of: 
❌ Loading entire "Market Research Methodology" document (5000 lines)
   into every conversation that might need it
You do: 
✅ SKILL.md (500 lines) serves as table of contents
   ↓
   Points to detailed files loaded ONLY when that specific
   aspect of research is needed:
   - competitive-analysis.md (loaded for competitor research)
   - customer-interviews.md (loaded for qual research)
   - survey-design.md (loaded for quant research)
Agents navigate skills like humans use a manual—they skim the overview, then go deep on the relevant section.

​
Skill Structure

​
Location

.claude/skills/{skill-name}/
Examples:
  • .claude/skills/conducting-market-research/
  • .claude/skills/writing-brand-consistent-content/
  • .claude/skills/managing-multi-step-projects/

​
Required: SKILL.md (Table of Contents)

Every skill must have a SKILL.md file that serves as the entry point. Rules:
  • Keep under 500 lines for optimal performance
  • Acts as an overview and navigation guide
  • Points to detailed files for specific aspects
  • Loaded into every conversation where this skill is used
Example: .claude/skills/conducting-market-research/SKILL.md 
# Conducting Market Research

## Overview
This skill provides methodologies for systematic market research including competitive analysis, customer insights, and market sizing.

## When to Use This Skill
- Analyzing competitor positioning and messaging
- Understanding customer needs and pain points
- Validating product-market fit
- Sizing market opportunities
- Identifying trends and shifts

## Core Principles
1. **Evidence-based**: All claims backed by data
2. **Systematic**: Follow established frameworks
3. **Triangulated**: Multiple sources validate findings
4. **Documented**: Clear audit trail of sources

## Research Types

### Competitive Analysis
For analyzing competitors, see: **competitive-analysis.md**

Use when you need to:
- Map competitive landscape
- Analyze competitor positioning
- Identify differentiation opportunities

### Customer Research
For understanding customers, see: **customer-research.md**

Use when you need to:
- Extract insights from interviews
- Analyze survey data
- Build personas

### Market Sizing
For quantifying opportunities, see: **market-sizing.md**

Use when you need to:
- Estimate TAM/SAM/SOM
- Forecast market growth
- Assess opportunity size

## Tools Used
- **Perplexity**: Web research, fact-checking
- **Firecrawl**: Competitor website analysis

## Output Standards
All research outputs must include:
1. Executive Summary (key insights)
2. Methodology (how research was conducted)
3. Findings (detailed analysis)
4. Sources (all data references)
5. Confidence Levels (high/medium/low for each claim)

## Related Skills
- **Analyzing Qualitative Data** (for coding interview transcripts)
- **Data Visualization** (for presenting findings)

---

*For specific research methodologies, see the detailed files above.*

​
Supporting Files (Loaded as Needed)

Example structure: 
.claude/skills/conducting-market-research/
├── SKILL.md (overview, always loaded)
├── competitive-analysis.md (loaded when analyzing competitors)
├── customer-research.md (loaded when researching customers)
├── market-sizing.md (loaded when sizing markets)
└── frameworks/
    ├── swot-analysis.md
    ├── positioning-map.md
    └── customer-interview-guide.md
When an agent needs to analyze a competitor:
  1. SKILL.md is loaded (overview)
  2. Agent sees “For competitive analysis, see competitive-analysis.md”
  3. competitive-analysis.md is loaded (detailed methodology)
  4. Agent executes research using that specific framework
Benefits:
  • ✅ Only relevant context is loaded (saves tokens)
  • ✅ Agent has detailed guidance when needed
  • ✅ Skill can be comprehensive without overwhelming every conversation

​
Skill Naming Convention

​
Use Gerund Form (Verb + -ing)

Good examples:
  • “Conducting Market Research”
  • “Writing Brand-Consistent Content”
  • “Analyzing Qualitative Data”
  • “Managing Multi-Step Projects”
  • “Creating Campaign Strategies”
Why gerund? It clearly describes the activity or capability the skill provides.

​
Acceptable Alternatives

Noun phrases:
  • “Market Research”
  • “Content Writing”
  • “Project Management”
Action-oriented:
  • “Conduct Market Research”
  • “Write Brand Content”

​
Avoid

❌ Vague names: “Helper”, “Utils”, “Tools” ❌ Overly generic: “Marketing”, “Content”, “Analysis” ❌ Inconsistent patterns: Mixing gerunds, nouns, and verbs confuses organization

​
Core Skills vs Domain Skills

​
Core Skills (Infrastructure Team Owns)

These ensure the system works consistently: 1. Orchestrating Projects
  • Multi-step coordination
  • Managing dependencies
  • Tracking progress (PLAN.md, TODO.md)
2. Ensuring Architectural Compliance
  • Brand consistency checks
  • File structure validation
  • Quality standards enforcement
3. Making Work Visible
  • Creating PLAN.md
  • Maintaining TODO.md
  • Progress reporting
Why infrastructure team owns these: They define HOW the architecture works. Marketing Architects can’t customize them without breaking the system.

​
Domain Skills (Marketing Architects Own)

These define YOUR specific capabilities: Examples:
  • Conducting Market Research (your methodologies)
  • Writing Brand-Consistent Content (your voice, frameworks)
  • Creating Campaign Strategies (your planning approach)
  • Analyzing Customer Data (your analysis frameworks)
Why marketing architects own these: They’re specific to YOUR brand, workflows, and expertise.

​
How Agents Use Skills

​
Skills are Mapped to Agents

Brand Analyst sub-agent
    ↓ has access to
Skills:
- Conducting Market Research
- Analyzing Qualitative Data
- Competitive Analysis

Content Writer sub-agent
    ↓ has access to
Skills:
- Writing Brand-Consistent Content
- Structuring Narratives
- Content Frameworks

Operations Manager
    ↓ has access to
Core Skills:
- Orchestrating Projects
- Ensuring Architectural Compliance
- Making Work Visible
Marketing Architects define which agents can use which skills in the agent definition files.

​
Invocation Flow

1. Agent receives task
   ↓
2. Agent checks: "Which skills do I have for this?"
   ↓
3. Agent loads SKILL.md (overview)
   ↓
4. Agent navigates to relevant section
   ↓
5. Agent loads detailed file if needed
   ↓
6. Agent executes using methodology
   ↓
7. Agent produces output following skill standards

​
Building a Skill: Step-by-Step

​
Step 1: Identify the Need

Ask: What capability do I need that I’m currently explaining ad-hoc? Examples:
  • “I keep explaining how to structure blog posts” → Create: “Writing Blog Posts” skill
  • “I’m repeatedly doing competitor analysis the same way” → Create: “Conducting Competitive Analysis” skill

​
Step 2: Define the Scope

What does this skill cover?
  • When should it be used?
  • What methodologies does it include?
  • What outputs does it produce?
  • Which agents will use it?

​
Step 3: Create the Directory Structure

.claude/skills/{skill-name}/
├── SKILL.md
└── [supporting files as needed]

​
Step 4: Write SKILL.md (Table of Contents)

Include:
  1. Overview - What this skill is
  2. When to Use - Specific scenarios
  3. Core Principles - Guiding philosophy
  4. Main Sections - Pointers to detailed files
  5. Tools Used - External tools leveraged
  6. Output Standards - What “done” looks like
  7. Related Skills - Connections to other skills
Keep under 500 lines.

​
Step 5: Create Supporting Files

For each major aspect of the skill:
  • Create a dedicated file (e.g., competitive-analysis.md)
  • Include detailed methodology
  • Provide examples and frameworks
  • Specify edge cases and variations

​
Step 6: Map to Agents

Update agent definitions to specify which agents can use this skill. 
# Brand Analyst Sub-agent

## Skills You Have Access To
- Conducting Market Research
- Analyzing Qualitative Data
- **[NEW] Competitive Analysis** ← Add here

​
Step 7: Test and Refine

Use the skill in real work:
  • Did the agent find what it needed?
  • Was the progressive disclosure effective?
  • Were outputs consistent and high-quality?
  • Did you need to add more detail anywhere?
Iterate based on usage.

​
Progressive Disclosure Patterns

​
Pattern 1: By Research Type

SKILL.md overview: 
## Research Types

### Competitive Analysis
See: competitive-analysis.md

### Customer Research
See: customer-research.md

### Market Sizing
See: market-sizing.md
Agent loads only the type it needs.

​
Pattern 2: By Content Format

SKILL.md overview: 
## Content Types

### Blog Posts
See: blog-posts.md

### Social Media
See: social-media.md

### Email Campaigns
See: email-campaigns.md

​
Pattern 3: By Complexity Level

SKILL.md overview: 
## Quick Reference (most common cases)
[Include 80% use cases here in SKILL.md]

## Advanced Scenarios
For complex cases, see: advanced-scenarios.md

​
Pattern 4: By Framework

SKILL.md overview: 
## Available Frameworks

### SWOT Analysis
See: frameworks/swot-analysis.md

### Positioning Map
See: frameworks/positioning-map.md

### Customer Journey Map
See: frameworks/customer-journey.md

​
Example: Complete Skill

Skill: Writing Brand-Consistent Content 
.claude/skills/writing-brand-consistent-content/
├── SKILL.md
├── voice-guidelines.md
├── content-types/
│   ├── blog-posts.md
│   ├── social-media.md
│   └── email.md
└── frameworks/
    ├── blog-post-structure.md
    ├── twitter-thread-structure.md
    └── email-sequence-structure.md
SKILL.md (excerpt): 
# Writing Brand-Consistent Content

## Overview
This skill ensures all content aligns with brand voice, messaging, and quality standards.

## Core Principle
**Every piece of content must:**
1. Reference /strategy/voice/ for tone
2. Reference /strategy/messaging/ for themes
3. Follow established content frameworks
4. Include audit trail (which strategy files used)

## Voice & Tone
For detailed voice guidelines, see: **voice-guidelines.md**

## Content Types

### Blog Posts
For blog post structure and best practices, see: **content-types/blog-posts.md**

### Social Media
For platform-specific social content, see: **content-types/social-media.md**

### Email Campaigns
For email writing guidelines, see: **content-types/email.md**

## Frameworks
Available structures:
- Blog Post Structure (frameworks/blog-post-structure.md)
- Twitter Thread Structure (frameworks/twitter-thread-structure.md)
- Email Sequence Structure (frameworks/email-sequence-structure.md)

## Quality Checklist
Before delivering content:
- [ ] Aligns with brand voice
- [ ] Uses messaging pillars
- [ ] Follows framework structure
- [ ] Includes relevant internal links
- [ ] No AI slop (generic, brand-less language)
When Content Writer creates a blog post:
  1. Loads SKILL.md (sees overview)
  2. Sees “For blog posts, see content-types/blog-posts.md”
  3. Loads blog-posts.md (detailed blog guidance)
  4. Sees “For structure, see frameworks/blog-post-structure.md”
  5. Loads blog-post-structure.md (exact framework)
  6. Writes blog post following all guidelines
Progressive disclosure at work—only loads what’s needed for THIS task.

​
Benefits of Well-Designed Skills

​
1. Consistency

Every time an agent uses the skill, it follows the same methodology. No variance, no forgetting steps.

​
2. Compounding Quality

Skills improve over time. Refine the skill once, all future work benefits.

​
3. Knowledge Capture

Your expertise is codified, not trapped in your head. The system becomes smarter.

​
4. Onboarding New Agents

Create a new sub-agent, give it access to existing skills, it’s instantly capable.

​
5. Scalability

Skills can be reused across projects, brands, contexts without re-explaining.

​
Common Questions

​
Q: How do I know if something should be a skill or part of strategy?

Ask:
  • Strategy = WHAT we say (voice, messaging, positioning)
  • Skill = HOW we do work (methodologies, frameworks)
Example:
  • “Our brand voice is professional yet approachable” → Strategy
  • “How to write in a professional yet approachable voice” → Skill

​
Q: Can skills reference other skills?

Yes! Include in “Related Skills” section of SKILL.md. Example: 
## Related Skills
This skill works well with:
- Analyzing Qualitative Data (for processing research)
- Content Frameworks (for structuring findings)

​
Q: What if a skill file grows beyond 500 lines?

Split it into supporting files. Before: 
SKILL.md (800 lines - too big!)
After: 
SKILL.md (400 lines - overview + pointers)
├── advanced-techniques.md (300 lines)
└── edge-cases.md (200 lines)

​
Q: Can multiple agents use the same skill?

Absolutely! That’s the point of reusability. Example: Both Brand Analyst and Campaign Strategist might use “Conducting Market Research” skill.

​
Q: Who decides which skills an agent can use?

Marketing Architect. You specify in the agent definition file.

​
What’s Next

Understand the final layer: Or see skills in action: Or dive deep into the principle:
“Skills turn expertise into infrastructure—build them once, benefit forever.”