our-claude-skills/docs/USER-GUIDE.md

OurDigital Custom Skills — User Guide

A comprehensive guide to the 52 custom skills in the our-claude-skills collection.

---

Table of Contents

1. Introduction
2. Quick Start
3. Skill Categories at a Glance
4. OurDigital Core (00-10)
5. SEO Tools (11-34)
6. Jamie Clinic (40-45)
7. NotebookLM Tools (50-53)
8. GTM / Analytics (60-62)
9. Notion Tools (31-32)
10. Reference Curator (90)
11. Multi-Agent Guide (91)
12. Architecture & Conventions
13. Creating New Skills
14. Troubleshooting

---

1. Introduction

What Are Custom Skills?

Custom skills are reusable, domain-specific instruction sets that extend Claude's capabilities. Each skill teaches Claude how to perform a particular task — writing a blog post in the OurDigital brand voice, running an SEO audit, managing a GTM container, or curating reference documentation.

Two Platforms, One Collection

Every skill is designed to work on two independent platforms:

Platform	Directive File	How It Runs	Best For
Claude Code (CLI)	code/CLAUDE.md	Direct Bash/Python execution, slash commands	Developers, automation, scripting
Claude Desktop (App)	desktop/SKILL.md	MCP tools only, conversational	Non-technical users, guided workflows

You do not need both. Pick the platform you use and follow the corresponding directive.

How Skills Are Invoked

Claude Code — Type a slash command in the CLI:

/seo-technical https://ourdigital.org
/reference-curator "MCP documentation" --depth deep
/notebooklm-studio generate audio "AI trends summary"

Claude Desktop — Use natural language triggers:

"Run a technical SEO audit on ourdigital.org"
"Create a podcast from my notebook"
"Write a blog post about AI marketing"

---

2. Quick Start

Prerequisites

Requirement	Used By	Install
Python 3.10+	SEO scripts, NotebookLM	brew install python
Node.js 18+	GTM scripts, crawlers	brew install node
Firecrawl MCP	Reference Curator, crawling	Configured in Claude Code MCP settings
Notion MCP	Notion skills	Configured in Claude Code MCP settings
Ahrefs MCP	SEO skills (11-34)	Configured in Claude Code MCP settings
notebooklm-py	NotebookLM skills (50-53)	pip install notebooklm-py && playwright install chromium

Installation

# Clone the repository
git clone https://github.com/ourdigital/our-claude-skills.git
cd our-claude-skills

# (Optional) Install SEO dependencies
pip install -r custom-skills/_ourdigital-shared/requirements/base.txt

# (Optional) Install Reference Curator
cd custom-skills/90-reference-curator && ./install.sh

# (Optional) Set up NotebookLM
pip install notebooklm-py
playwright install chromium
notebooklm login

Verifying Slash Commands

Once the repository is cloned and opened in Claude Code, slash commands are automatically available from .claude/commands/. Type / in Claude Code to see the full list.

---

3. Skill Categories at a Glance

Range	Domain	Count	Description
00	Settings & Meta	1	Claude Code settings optimization
01-10	OurDigital Core	10	Brand, content, design, business tools
11-34	SEO	24	Technical audit through executive reporting
40-45	Jamie Clinic	5	Medical brand content and social media
50-53	NotebookLM	4	AI notebook management and content generation
60-62	GTM / Analytics	3	Tag management and monitoring
31-32	Notion	2	Workspace organization and writing
90	Reference Curator	1 (6 sub-skills)	Documentation curation pipeline
91	Multi-Agent	1	Agent collaboration framework

Total: 52 skills (including Reference Curator sub-skills)

---

4. OurDigital Core (00-10)

Overview

The core skills handle brand identity, content creation across multiple channels, visual design, advertising, training, and business operations.

00 — Settings Audit

Purpose	Optimize Claude Code settings, audit token usage, diagnose MCP errors
Command	/settings-audit
Triggers	"settings audit", "exceed response limit", "MCP error"
Platforms	Code + Desktop

What it does:

• Analyzes ~/.claude/settings.json for optimization opportunities
• Reports token usage patterns
• Diagnoses MCP server connection issues
• Recommends permission and hook configurations

Example:

/settings-audit

01 — Brand Guide

Purpose	OurDigital brand standards, writing style, visual identity reference
Command	/ourdigital-brand-guide
Triggers	"brand guide", "writing style", "brand standards"
Platforms	Desktop (slash command adapted from desktop)

What it does:

• Provides the OurDigital "Digital Clinic" brand metaphor guidelines
• Defines tone and voice per channel (ourdigital.org, blog, journal, ourstory.day)
• Specifies color palette, typography (Noto Sans KR, Inter), and visual rules
• Enforces the three brand values: Data-driven, In-action, Marketing Science

Brand channels:

Channel	Domain	Tone
Main Site	ourdigital.org	Professional, data-driven
Blog	blog.ourdigital.org	Analytical, educational
Journal	journal.ourdigital.org	Conversational, poetic
OurStory	ourstory.day	Intimate, authentic
Corporate	dintelligence.co.kr	Professional, B2B

02 — Blog Writer

Purpose	Korean blog creation for blog.ourdigital.org (Ghost)
Triggers	"write blog post", "블로그 작성"
Platforms	Desktop

What it does:

• Writes Korean-language blog posts following OurDigital editorial standards
• Structures content with proper headings, subheadings, and CTA
• Applies the "Digital Clinic" analytical tone
• Targets 800-1,200 words per post

03 — Journal Writer

Purpose	English essay/article for journal.ourdigital.org (Ghost)
Triggers	"write essay", "journal article"
Platforms	Desktop

What it does:

• Creates English-language essays and long-form articles
• Uses the conversational, poetic tone defined for the Journal channel
• Handles personal narratives and cultural commentary

04 — Research

Purpose	Research-to-Blog workflow
Command	/ourdigital-research
Triggers	"research this", "blog post research"
Platforms	Code + Desktop

What it does:

• Conducts web research on a given topic
• Synthesizes findings into a structured outline
• Optionally transitions into blog/journal writing with the appropriate skill

05 — Document

Purpose	Notion-to-Presentation workflow
Command	/ourdigital-presentation
Triggers	"create document", "Notion export", "presentation"
Platforms	Code + Desktop

What it does:

• Exports Notion content into presentation-ready format
• Structures slides with OurDigital brand visual identity
• Supports Korean and English bilingual content

06 — Designer

Purpose	Visual storytelling and image prompt generation
Command	/ourdigital-designer
Triggers	"create image prompt", "visual design"
Platforms	Code + Desktop

What it does:

• Generates detailed image prompts for AI image generators
• Follows OurDigital visual brand guidelines (color palette, typography)
• Creates visual storytelling concepts for blog posts and social media

07 — Ad Manager

Purpose	Ad copywriting and keyword research
Triggers	"ad copy", "keyword research", "광고 카피"
Platforms	Desktop

What it does:

• Writes ad copy for Google Ads, Meta Ads, and Naver Ads
• Conducts keyword research for paid campaigns
• Generates A/B test variants
• Supports Korean and English ad copy

08 — Trainer

Purpose	Training material creation
Triggers	"create training", "교육 자료"
Platforms	Desktop

What it does:

• Creates structured training materials and curricula
• Generates quizzes, exercises, and assessment rubrics
• Supports bilingual (Korean/English) training content

09 — Back Office

Purpose	Business document creation
Triggers	"create proposal", "견적서", "invoice"
Platforms	Desktop

What it does:

• Generates business documents: proposals, invoices, contracts
• Applies OurDigital corporate branding
• Supports Korean business document conventions

10 — Skill Creator

Purpose	Meta skill for creating new skills
Triggers	"create skill", "init skill"
Platforms	Desktop

What it does:

• Scaffolds new skill directory structure
• Generates boilerplate CLAUDE.md and SKILL.md files
• Follows the dual-platform convention automatically
• Assigns the next available skill number

Example:

python example-skills/skills-main/skill-creator/scripts/init_skill.py my-new-skill --path custom-skills/

---

5. SEO Tools (11-34)

Overview

The SEO suite is the largest skill group, covering everything from technical crawlability to executive dashboards. Skills 11-18 focus on site-level audits, 19-28 on strategy and intelligence, and 29-34 on specialized operations and reporting.

All SEO skills integrate with:

• Ahrefs MCP — Domain metrics, backlinks, keyword data
• Notion — SEO Audit Log database for persistent records
• Python scripts — Shared base_client.py patterns (RateLimiter, ConfigManager, BaseAsyncClient)

Audit & Technical (11-18)

11 — Comprehensive Audit

Purpose	Unified 6-stage SEO audit orchestrator
Triggers	"comprehensive SEO", "full SEO audit"
Platforms	Code + Desktop

What it does: Orchestrates skills 12-18 into a single comprehensive audit. Runs technical, on-page, Core Web Vitals, Search Console, schema, and local audits in sequence, producing a unified report.

12 — Technical Audit

Purpose	Robots.txt, sitemap, crawlability analysis
Command	/seo-technical
Triggers	"technical SEO", "crawlability"
Platforms	Code + Desktop

What it does:

• Validates robots.txt directives
• Checks XML sitemap completeness and structure
• Analyzes crawl errors and redirect chains
• Tests HTTPS configuration and canonical tags

Example:

/seo-technical https://ourdigital.org

13 — On-Page Audit

Purpose	Meta tags, headings, internal links
Command	/seo-on-page
Triggers	"on-page SEO", "meta tags"
Platforms	Code + Desktop

What it does:

• Analyzes title tags, meta descriptions, heading hierarchy
• Checks internal linking structure and anchor text
• Validates Open Graph and Twitter Card tags
• Reviews content quality signals

14 — Core Web Vitals

Purpose	LCP, CLS, FID, INP performance metrics
Command	/seo-vitals
Triggers	"Core Web Vitals", "page speed"
Platforms	Code + Desktop

What it does:

• Measures Largest Contentful Paint (LCP)
• Measures Cumulative Layout Shift (CLS)
• Measures First Input Delay (FID) / Interaction to Next Paint (INP)
• Provides actionable optimization recommendations

15 — Search Console

Purpose	Google Search Console data analysis
Command	/seo-gsc
Triggers	"Search Console", "GSC", "rankings"
Platforms	Code + Desktop

What it does:

• Analyzes search performance data (clicks, impressions, CTR, position)
• Identifies keyword opportunities and declining pages
• Reports on index coverage issues
• Tracks ranking changes over time

16 — Schema Validator

Purpose	Structured data validation
Command	/seo-schema-validator
Triggers	"validate schema", "JSON-LD check"
Platforms	Code + Desktop

What it does:

• Validates existing JSON-LD, Microdata, and RDFa markup
• Checks against Google's rich result requirements
• Reports missing required and recommended properties
• Tests against Schema.org vocabulary

17 — Schema Generator

Purpose	Structured data creation
Command	/seo-schema-generator
Triggers	"generate schema", "create JSON-LD"
Platforms	Code + Desktop

What it does:

• Generates JSON-LD markup for various schema types
• Supports: Article, LocalBusiness, Product, FAQ, HowTo, Organization, BreadcrumbList
• Validates output before delivery
• Provides implementation instructions

18 — Local Audit

Purpose	Local SEO: NAP consistency, GBP, citations
Command	/seo-local
Triggers	"local SEO", "Google Business Profile"
Platforms	Code + Desktop

What it does:

• Checks Name-Address-Phone (NAP) consistency across the web
• Audits Google Business Profile completeness
• Analyzes local citation sources
• Reviews local keyword targeting

Strategy & Intelligence (19-28)

19 — Keyword Strategy

Purpose	Keyword expansion, intent classification, clustering, gap analysis
Command	/seo-keyword-strategy
Triggers	"keyword research", "keyword strategy"
Platforms	Code + Desktop

What it does:

• Expands seed keywords using Ahrefs data
• Classifies search intent (informational, navigational, commercial, transactional)
• Clusters keywords by topic and SERP similarity
• Identifies content gaps vs. competitors

Example:

/seo-keyword-strategy "digital marketing Seoul" --competitors competitor1.com,competitor2.com

20 — SERP Analysis

Purpose	Google/Naver SERP features and competitor positions
Command	/seo-serp-analysis
Triggers	"SERP analysis", "SERP features"
Platforms	Code + Desktop

What it does:

• Detects SERP features (featured snippets, PAA, knowledge panels, local packs)
• Analyzes competitor positioning for target keywords
• Supports both Google and Naver (Korean search engine)
• Identifies SERP feature opportunities

21 — Position Tracking

Purpose	Rank monitoring, visibility scores, alerts
Command	/seo-position-tracking
Triggers	"rank tracking", "position monitoring"
Platforms	Code + Desktop

What it does:

• Monitors keyword rankings over time
• Calculates visibility scores and share of voice
• Generates alerts for significant ranking changes
• Tracks competitor movements

22 — Link Building

Purpose	Backlink audit, toxic link detection, link gap analysis
Command	/seo-link-building
Triggers	"backlink audit", "link building"
Platforms	Code + Desktop

What it does:

• Audits existing backlink profile using Ahrefs data
• Identifies toxic or spammy links for disavow
• Finds link gap opportunities vs. competitors
• Suggests outreach targets and link-worthy content

23 — Content Strategy

Purpose	Content audit, decay detection, content briefs, topic clusters
Command	/seo-content-strategy
Triggers	"content strategy", "content audit"
Platforms	Code + Desktop

What it does:

• Audits existing content inventory and performance
• Detects content decay (declining traffic/rankings)
• Generates content briefs with keyword targets
• Maps topic clusters and internal linking strategy

24 — E-commerce SEO

Purpose	Product page audit, product schema validation
Command	/seo-ecommerce
Triggers	"e-commerce SEO", "product SEO"
Platforms	Code + Desktop

What it does:

• Audits product page SEO elements (titles, descriptions, images)
• Validates Product schema markup
• Checks category page optimization
• Reviews faceted navigation and crawl efficiency

25 — KPI Framework

Purpose	Unified SEO KPIs, health scores, ROI calculation
Command	/seo-kpi-framework
Triggers	"SEO KPI", "SEO performance"
Platforms	Code + Desktop

What it does:

• Defines and tracks unified SEO KPIs across all audit dimensions
• Calculates composite health scores
• Estimates SEO ROI based on traffic and conversion data
• Generates KPI dashboards for stakeholder reporting

26 — International SEO

Purpose	Hreflang validation, content parity, multi-language analysis
Command	/seo-international
Triggers	"international SEO", "hreflang"
Platforms	Code + Desktop

What it does:

• Validates hreflang tag implementation
• Checks content parity across language versions
• Analyzes international URL structure (subdomain vs. subdirectory vs. ccTLD)
• Reviews geo-targeting settings

27 — AI Visibility

Purpose	AI search citations, brand radar, share of voice in AI
Command	/seo-ai-visibility
Triggers	"AI visibility", "AI search"
Platforms	Code + Desktop

What it does:

• Monitors brand mentions in AI-generated search results
• Tracks share of voice in AI search (ChatGPT, Perplexity, Google AI Overview)
• Analyzes citation patterns and cited domains
• Measures AI impression history and trends

28 — Knowledge Graph

Purpose	Entity SEO, Knowledge Panel optimization, PAA analysis
Command	/seo-knowledge-graph
Triggers	"knowledge graph", "entity SEO"
Platforms	Code + Desktop

What it does:

• Audits entity presence in Google's Knowledge Graph
• Optimizes for Knowledge Panel appearance
• Analyzes "People Also Ask" opportunities
• Reviews entity disambiguation and structured data connections

Specialized Operations (29-34)

29 — Gateway Architect

Purpose	Gateway page strategy and architecture design
Command	/seo-gateway-architect
Triggers	"SEO strategy", "gateway pages"
Platforms	Code + Desktop

What it does:

• Designs gateway page architecture for target keyword clusters
• Maps URL structures and internal linking
• Plans content hierarchy and pillar-cluster relationships

30 — Gateway Builder

Purpose	Gateway page content generation
Command	/seo-gateway-builder
Triggers	"build gateway page"
Platforms	Code + Desktop

What it does:

• Generates optimized gateway page content based on architect blueprints
• Includes meta tags, headings, body copy, and internal links
• Follows SEO best practices for content structure

31 — Competitor Intel

Purpose	Competitor profiling, benchmarking, threat analysis
Command	/seo-competitor-intel
Triggers	"competitor analysis", "competitive intel"
Platforms	Code + Desktop

What it does:

• Profiles competitor SEO strategies using Ahrefs data
• Benchmarks domain authority, traffic, and keyword overlap
• Identifies competitive threats and opportunities
• Generates comparative reports

32 — Crawl Budget

Purpose	Server log analysis, bot profiling, crawl waste identification
Command	/seo-crawl-budget
Triggers	"crawl budget", "log analysis"
Platforms	Code + Desktop

What it does:

• Analyzes server access logs for bot crawl patterns
• Profiles Googlebot, Bingbot, and other crawler behavior
• Identifies crawl waste (duplicate content, soft 404s, parameter URLs)
• Recommends crawl budget optimization strategies

33 — Migration Planner

Purpose	Site migration planning, redirect mapping, risk assessment
Command	/seo-migration-planner
Triggers	"site migration", "domain move", "사이트 이전"
Platforms	Code + Desktop

What it does:

• Creates pre-migration baseline snapshots
• Generates redirect mapping (old URL → new URL)
• Assesses migration risk factors
• Plans post-migration monitoring checkpoints
• Supports domain moves, CMS migrations, HTTPS transitions, and URL restructuring

34 — Reporting Dashboard

Purpose	Executive reports, HTML dashboards, data aggregation
Command	/seo-reporting-dashboard
Triggers	"SEO report", "SEO dashboard", "보고서"
Platforms	Code + Desktop

What it does:

• Aggregates data from all other SEO skills into unified reports
• Generates executive summaries with key metrics and trends
• Creates interactive HTML dashboards
• Supports Korean-language report summaries
• Exports to multiple formats (Markdown, HTML, PDF-ready)

---

6. Jamie Clinic (40-45)

Overview

Skills for Jamie Plastic Surgery Clinic's digital presence. All content must follow strict medical brand compliance guidelines and supports Korean-language content.

40 — Brand Editor

Purpose	Jamie Clinic branded content generation
Command	/jamie-editor
Triggers	"write Jamie blog", "Jamie content"
Platforms	Code + Desktop

What it does:

• Generates blog posts and marketing content in Jamie's brand voice
• Follows medical content compliance guidelines
• Maintains consistent tone: professional yet approachable
• Supports Korean and English content

41 — Brand Audit

Purpose	Content review and brand compliance checking
Command	/jamie-audit
Triggers	"review content", "brand audit"
Platforms	Code + Desktop

What it does:

• Reviews existing content for brand consistency
• Checks medical claims against compliance standards
• Validates tone, terminology, and visual brand adherence
• Generates compliance reports with specific revision suggestions

43 — YouTube Manager

Purpose	YouTube channel SEO audit and content optimization
Command	N/A (use trigger)
Triggers	"YouTube SEO", "YT optimization"
Platforms	Code + Desktop

What it does:

• Audits YouTube channel and video SEO
• Optimizes titles, descriptions, tags, and thumbnails
• Analyzes competitor YouTube channels
• Suggests content strategy for video SEO

44 — Subtitle Checker

Purpose	YouTube subtitle file (SBV) validation and typo correction
Command	N/A (use trigger)
Triggers	"check subtitles", "subtitle QA"
Platforms	Code + Desktop

What it does:

• Validates SBV subtitle file format and timing
• Detects and corrects Korean/English typos
• Checks subtitle synchronization
• Ensures accessibility compliance

45 — Instagram Manager

Purpose	Instagram account management and strategy
Command	N/A (use trigger)
Triggers	"Instagram management", "IG strategy"
Platforms	Code + Desktop

What it does:

• Plans Instagram content calendar
• Generates caption copy and hashtag strategies
• Analyzes engagement metrics
• Maintains brand visual consistency on the platform

---

7. NotebookLM Tools (50-53)

Overview

Four specialized skills for Google's NotebookLM platform, powered by the notebooklm-py CLI.

Prerequisites (all skills):

pip install notebooklm-py
playwright install chromium
notebooklm login    # Browser-based Google OAuth

50 — Agent

Purpose	Q&A from notebooks with Gemini-powered citations
Command	/notebooklm-agent
Triggers	"ask NotebookLM", "query notebook", "노트북 질문"
Platforms	Code + Desktop

What it does:

• Queries NotebookLM notebooks with natural language questions
• Returns answers with source citations
• Supports JSON output for programmatic use
• Can target specific notebooks by name or ID

Example:

notebooklm ask "What are the key findings about AI adoption?" --json

51 — Automation

Purpose	Programmatic notebook, source, and artifact management
Command	/notebooklm-automation
Triggers	"create notebook", "add sources", "노트북 관리"
Platforms	Code + Desktop

What it does:

• Creates and manages notebooks programmatically
• Adds sources (URLs, documents, text) to notebooks
• Lists and manages existing notebooks and sources
• Handles batch operations

Example:

notebooklm create "Research Project Alpha"
notebooklm source add "https://example.com/paper.pdf"
notebooklm list --json

52 — Studio

Purpose	Content generation: audio, video, quizzes, flashcards, slides, mind maps
Command	/notebooklm-studio
Triggers	"create podcast", "generate video", "make quiz", "팟캐스트 만들기"
Platforms	Code + Desktop

What it does:

• Generates audio podcasts (deep-dive, brief, critique, debate formats)
• Creates videos (classic, whiteboard, kawaii, anime, pixel, and more styles)
• Builds quizzes and flashcards from notebook content
• Generates slide decks, infographics, and mind maps

Generation times:

Content Type	Approximate Time
Mind map	Instant
Quiz/Flashcards	5-15 min
Audio podcast	10-20 min
Video	15-45 min

Example:

notebooklm generate audio "Create a debate about AI ethics" --format debate
notebooklm generate video --style whiteboard
notebooklm download audio ./podcast.mp3

53 — Research

Purpose	Web and Google Drive research, source discovery
Command	/notebooklm-research
Triggers	"research topic", "find sources", "리서치"
Platforms	Code + Desktop

What it does:

• Conducts web research and auto-imports results into notebooks
• Searches Google Drive for relevant documents
• Supports fast mode (5-10 sources, seconds) and deep mode (20+ sources, 2-5 min)
• Non-blocking operation with status polling

Example:

# Fast web research
notebooklm source add-research "machine learning trends"

# Deep research with auto-import
notebooklm source add-research "AI regulation" --mode deep --import-all

# Drive research
notebooklm source add-research "quarterly report" --from drive

---

8. GTM / Analytics (60-62)

Overview

Google Tag Manager skills for auditing, managing, and monitoring tag implementations.

60 — GTM Audit

Purpose	GTM container audit and analysis
Command	/gtm-audit
Triggers	"audit GTM", "GTM analysis"
Platforms	Code + Desktop

What it does:

• Audits GTM container configuration
• Identifies unused or broken tags
• Checks trigger configurations and firing rules
• Reviews variable definitions and data layer usage
• Generates audit report with recommendations

61 — GTM Manager

Purpose	GTM management and dataLayer operations
Command	/gtm-manager
Triggers	"GTM manager", "dataLayer"
Platforms	Code

What it does:

• Manages GTM tags, triggers, and variables
• Configures dataLayer push events
• Creates and modifies GTM workspace configurations
• Supports batch tag operations

62 — GTM Guardian

Purpose	Live tag monitoring and validation
Command	N/A (use trigger)
Triggers	"monitor GTM", "tag validation"
Platforms	Code + Desktop

What it does:

• Monitors live tag firing in real-time
• Validates tag configurations against best practices
• Alerts on tag errors, slow-firing tags, or missing data
• Acts as a knowledge assistant for GTM questions

---

9. Notion Tools (31-32)

Overview

Skills for managing and writing content within Notion workspaces. Requires the Notion MCP integration.

31 — Notion Organizer

Purpose	Notion workspace management and organization
Command	/notion-organizer
Triggers	"organize Notion", "workspace cleanup"
Platforms	Code + Desktop

What it does:

• Organizes Notion pages and databases
• Cleans up workspace structure
• Manages page properties and relations
• Supports batch operations on Notion content

32 — Notion Writer

Purpose	Push markdown content to Notion pages or databases
Command	/notion-writer
Triggers	"write to Notion", "export to Notion"
Platforms	Code + Desktop

What it does:

• Converts markdown content to Notion blocks
• Creates new pages in specified databases
• Updates existing page content
• Preserves formatting, headings, code blocks, and lists

---

10. Reference Curator (90)

Overview

A modular pipeline for curating, processing, and exporting reference documentation from the web. Consists of 6 sub-skills orchestrated by a single pipeline command.

Pipeline Command

Command	/reference-curator
Purpose	Full pipeline: discover → crawl → store → distill → review → export

Input Modes

Mode	Example	Pipeline Start
Topic	"Claude system prompts"	Discovery (Stage 1)
URLs	https://docs.anthropic.com/...	Crawler (Stage 2, skip discovery)
Manifest	./manifest.json	Crawler (resume)

Key Arguments

Argument	Default	Description
--depth	standard	Crawl depth: light, standard, deep, full
--output	~/Documents/reference-library/	Output directory (asks permission to create)
--max-sources	100	Max sources to discover
--max-pages	varies by depth	Max pages per source
--auto-approve	off	Auto-approve documents scoring above threshold
--threshold	0.85	QA approval threshold
--export-format	project_files	Output format: project_files, fine_tuning, jsonl

Depth Levels

light       ████░░░░░░░░░░░░  ~20 pages   Main content only      Fastest
standard    ████████░░░░░░░░  ~50 pages   Main + links           Fast
deep        ████████████░░░░  ~150 pages  Full page + HTML       Moderate
full        ████████████████  ~500 pages  Everything + raw HTML  Slow

Pipeline Stages

1. reference-discovery     Search & validate authoritative sources
2. web-crawler             Crawl pages using Firecrawl (depth-controlled)
3. content-repository      Store documents with version tracking
4. content-distiller       Summarize and extract key concepts
        ↕ QA Loop
5. quality-reviewer        Score, approve, refactor, or reject
6. markdown-exporter       Export to final format

Output Structure

~/Documents/reference-library/
├── {topic-slug}/
│   ├── README.md                   # Index with table of contents
│   ├── 00-page-name.md             # Individual page files
│   ├── 01-page-name.md
│   ├── {topic-slug}-complete.md    # Combined bundle
│   └── manifest.json               # Crawl metadata
├── pipeline_state/                 # Resume state
└── exports/                        # Fine-tuning / JSONL exports

Examples

# Quick scan of a single doc page
/reference-curator https://docs.example.com/api --depth light

# Standard documentation crawl
/reference-curator "Glossary for WordPress" --max-sources 5

# Deep crawl with full page content
/reference-curator https://docs.codeat.co/glossary/ --depth deep

# Full archival crawl
/reference-curator https://docs.anthropic.com --depth full --max-pages 300

# Auto-approve with fine-tuning export
/reference-curator "MCP servers" --depth deep --auto-approve --export-format fine_tuning

Sub-skills (Individual Use)

Sub-skill	Command	Purpose
reference-discovery	/reference-discovery	Search & validate sources
web-crawler	/web-crawler	Multi-backend crawling
content-repository	/content-repository	MySQL storage with versioning
content-distiller	/content-distiller	Summarization & extraction
quality-reviewer	/quality-reviewer	QA loop routing
markdown-exporter	/markdown-exporter	Export to markdown/JSONL

---

11. Multi-Agent Guide (91)

Overview

Sets up a multi-agent collaboration framework for projects where Claude Code, Gemini, Codex, and human reviewers work together.

Command	/multi-agent-guide
Triggers	"set up multi-agent", "agent guardrails"

Agent Hierarchy

            ┌─────────────────┐
            │   Claude Code   │
            │  (Lead Agent)   │
            └────────┬────────┘
                     │
      ┌──────────────┼──────────────┐
      v              v              v
┌──────────┐  ┌──────────┐  ┌──────────┐
│  Gemini  │  │  Codex   │  │  Human   │
│(Research)│  │ (Speed)  │  │ (Review) │
└──────────┘  └──────────┘  └──────────┘

What It Creates

your-project/
├── .agent-state/
│   ├── tasks.yaml          # Task registry
│   └── locks.yaml          # Lock registry
├── tools/
│   └── check-ownership.py  # Ownership verification
├── MULTI_AGENT_FRAMEWORK.md  # Consolidated rules
├── GEMINI.md               # Gemini agent directive
└── CODEX.md                # Codex agent directive

Commit Convention

[Agent] type(scope): description

[Claude] feat(core): implement authentication
[Gemini] docs(api): update API documentation
[Codex] test(models): add unit tests

---

12. Architecture & Conventions

Dual-Platform Structure

Every skill follows a consistent directory structure:

XX-skill-name/
├── code/                    # Claude Code version
│   ├── CLAUDE.md            # Action-oriented directive
│   ├── scripts/             # Executable Python/Bash
│   └── docs/                # Documentation
│
├── desktop/                 # Claude Desktop version
│   ├── SKILL.md             # Skill directive with YAML frontmatter
│   ├── skill.yaml           # Extended metadata (optional)
│   └── tools/               # MCP tool documentation
│
├── shared/                  # Shared resources (optional)
│   ├── references/          # Brand guidelines, data files
│   ├── templates/           # Output templates
│   └── scripts/             # Shared scripts
│
└── README.md                # Skill overview

Platform Differences

Aspect	code/ (Claude Code)	desktop/ (Claude Desktop)
Directive	CLAUDE.md	SKILL.md (YAML frontmatter)
Execution	Direct Bash/Python	MCP tools only
Scripts	Required for automation	Reference only
Slash commands	.claude/commands/*.md	N/A

Shared Configuration

All skills share common configuration at:

custom-skills/_ourdigital-shared/
├── config/
│   └── ourdigital.yaml      # Brand, channels, colors, Notion, exports
├── requirements/
│   └── base.txt             # Python dependencies
└── installer/               # Shared setup scripts

Skill Numbering Convention

Range	Domain	Status
00	Settings & Meta	Active
01-09	OurDigital Core	Active
10	Skill Creator	Active
11-34	SEO Tools	Active
35-39	SEO (Reserved)	Available
40-49	Jamie Clinic	Active (40-45)
50-59	NotebookLM	Active (50-53)
60-69	GTM/Analytics	Active (60-62)
70-89	Unassigned	Available
90-99	Infrastructure	Active (90-91)

Design Principles

1. One thing done well — Each skill focuses on a single capability
2. Directives under 1,500 words — Concise and actionable
3. Self-contained — Each platform version is fully independent
4. Code-first — Build the Claude Code version first, then Desktop
5. Progressive numbering — Logical grouping by domain

---

13. Creating New Skills

Using the Skill Creator

python example-skills/skills-main/skill-creator/scripts/init_skill.py my-new-skill --path custom-skills/

Manual Creation

1. Choose a number from the available ranges (see numbering convention above)
2. Create the directory structure:

   mkdir -p custom-skills/XX-my-skill/{code,desktop,shared}
   

3. Write code/CLAUDE.md — Action-oriented directive for Claude Code
4. Write desktop/SKILL.md — Conversational directive with YAML frontmatter
5. Add scripts if needed in code/scripts/
6. Create a slash command in .claude/commands/my-skill.md
7. Update CLAUDE.md at the repo root to include your skill in the summary table

SKILL.md Frontmatter Format

---
name: my-skill-name
description: |
  Brief description of what the skill does.
  Triggers: keyword1, keyword2, 한국어 트리거.
---

# Skill Title

Directive content starts here...

Slash Command Format

---
description: Short description shown in command list
argument-hint: <required-arg> [--optional flag]
allowed-tools: Read, Write, Bash, WebSearch, WebFetch
---

# Skill Title

Instructions for Claude Code when this command is invoked...

---

14. Troubleshooting

Common Issues

Slash commands not showing up:

• Ensure the repo is opened as the working directory in Claude Code
• Check that .claude/commands/*.md files have valid YAML frontmatter
• Restart Claude Code if commands were recently added

MCP tool errors:

• Run /settings-audit to diagnose MCP connection issues
• Verify MCP servers are configured in Claude Code settings
• Check that API keys are set in environment variables

NotebookLM authentication:

notebooklm login    # Re-authenticate if session expired

SEO skill Python errors:

pip install -r custom-skills/_ourdigital-shared/requirements/base.txt

Reference Curator output directory:

• Default output: ~/Documents/reference-library/
• The pipeline will ask permission before creating new directories
• Never stores output in skill-related directories

Getting Help

• Repository: https://github.com/ourdigital/our-claude-skills
• Issues: https://github.com/ourdigital/our-claude-skills/issues
• Run /settings-audit for self-diagnostics

---

Last updated: 2026-02-24