What makes a skill different?
Consistent Quality. Every Time.
- →Starts from zero every time
- →You manually paste context (or forget to)
- →No access to your docs, examples, or tools
- →Quality varies based on what you remember to include
- →Pulls from 12+ reference docs (ICP, voice, past examples)
- →Dynamically loads only what this task needs
- →Orchestrates 15+ refined prompts in sequence
- →Uses tools: reads files, searches context, calls agents
- →Validates output automatically
You run one command. The skill handles the rest.
The Context Loading Pattern
Run /produce-content. Watch 5 skills coordinate automatically:
CONTEXT LOADING (automatic)
- ├── ICP synthesis → Who you're writing for
- ├── Voice & tone standards → How you sound
- ├── Content principles → What "good" looks like
- ├── Content type mapping → Matching format to what you're creating
- └── Recent outputs → Pattern consistency
EXECUTION (orchestrates skills)
- ├── Calls /persuasive-copywriting → Masters framework applied
- ├── Calls /generate-image → Visuals generated
- ├── Runs internal pipeline:
- │ └── Angles → Outline → Draft → Score → Polish
- └── All steps use context from Step 1
QUALITY GATES (automatic validation)
- ├── Calls /edit-content → Editorial polish
- ├── Calls /skeptical-buyer → Buyer lens check
- ├── Calls /final-content-reviewer → Quality gate
- └── Voice consistency verified
OUTPUT: Blog + LinkedIn + Newsletter + Images
1 command → 5 skills → 17 steps handled
Week 1: Magic.
“Wait, that actually worked?”
Week 4: Normal.
“I can't imagine going back to the old way.”
Week 1: Magic.
“Wait, that actually worked?”
Week 4: Normal.
“I can't imagine going back to the old way.”
Deep Planning: The Meta-Orchestrator
This skill instantly makes you a pro user of AI. Full stop.
# Deep Planning Repository-aware planning methodology for complex tasks.Covers the full lifecycle: context gathering → plan creation→ execution with checkpoints → completion. ## When to Use This Skill **Apply when:**├── Planning complex, multi-step tasks (3+ phases)├── Building prompts or specifications for agents├── Needing to gather context from the repository├── Decomposing work into actionable steps└── Work that may span multiple sessions **Do NOT use for:**├── Simple single-step tasks (just do them)├── Tasks with existing specs (use `specification-driven-development`)├── Pure research/exploration (use Explore agent)└── Software architecture (use `feature-planning-and-decomposition`)The 7 Phases
Workstream Alignment
Which mental mode?
Context Gathering
Find before create
Plan Creation
EnterPlanMode decision
Execution Setup
Tools, checkpoints, limits
Execution
With phase checkpoints
Re-Plan or Continue
Evaluation gates
Completion
Summary, learnings, cleanup
Deep planning doesn't just plan. It knows which skills and agents to call for each step. That means zero guessing about what to do next.
ICP Builder: 9 Agents, Zero Guessing
From raw evidence to scored accounts. Not a template. A system.
# ICP Builder 9-agent pipeline that constructs your Ideal CustomerProfile from evidence — not assumptions. ## Pipeline Architecture Phase 1 — Evidence Collection (parallel):├── Agent 01: Market evidence (web + enrichment)├── Agent 02: Customer intelligence (CRM mining)└── Agent 03: Competitive landscape (category warfare) Phase 2 — Synthesis:├── Agent 04: Pain-Qualification-Signal synthesis└── Agent 05: Persona hypothesis generation ── Human Review Gate ── Phase 3 — Downstream Construction:├── Agent 06: Scoring model (4-dimension weights)└── Agent 07: Signal playbook with triggered actions ## Lifecycle Commands /icp build → Full 9-agent construction/icp discover → Find matching companies/icp score-batch → Score and tier accounts/icp deep-dive → Research T1 accountsThe 9 Agents
Market Evidence Collector
Web + enrichment data
Customer Intelligence Miner
CRM patterns + pain signals
Competitive Landscape Mapper
Category warfare clustering
PQS Synthesizer
Pain → Qualification → Signal
Persona Hypothesis Generator
Evidence-based buyer profiles
Human Review Gate
You validate before scoring
Scoring Model Builder
4-dimension weighted formula
Signal Playbook Architect
Intent signals → triggered actions
Upstream agents collect evidence. Downstream agents score it. The human review gate keeps you in control. That's how you build an ICP you can trust.
Meeting Prep: 8 Types, Zero Generic Prep
Same skill, different depth. A board meeting gets governance context. A coffee chat gets conversation starters.
# Meeting Prep Context-adaptive dossier generator. One command.Eight meeting types. Research depth that matches. ## Meeting Type Router /meeting-prep "Discovery call with Acme Corp"→ Type: Sales Discovery→ Depth: FULL (company + attendees + MEDDPICC)→ Modules: research_prospect, pain_investigation, tech_stack_discovery, icp_classifier /meeting-prep "Coffee with Sarah from Stripe"→ Type: Informational→ Depth: LIGHT (highlights + connection points)→ Modules: attendee_research (tier 2) /meeting-prep "Board meeting for CLAL"→ Type: Board/Advisory→ Depth: FULL (backgrounds + governance)→ Modules: attendee_research (tier 1), prior_interactions ## Evidence Standards [VERIFIED] — Found in reliable source[INFERRED] — Deduced from signals[NOT_FOUND] — Searched, couldn't determineThe 7-Step Workflow
Gather Meeting Context
Names, companies, your objectives
Classify Meeting Type
Auto-detect from 8 types
Route Research Modules
Activate depth-appropriate modules
Execute Research
WebSearch + repo intel in parallel
Handle Unknowns
Graceful degradation, never fabricate
Synthesize Dossier
One scannable markdown artifact
Present to User
3-bullet exec summary + full dossier
A sales discovery gets MEDDPICC-level intelligence. A coffee chat gets conversation starters. The skill reads the context so you don't prep the wrong way.
How a skill actually works.
Load. Apply. Refine. Ship.
.claude/skills/copywriting/SKILL.md ## Skill Loaded: Copywriting **Type:** Short-form, conversion-focused copy **Formats recognized:**├── Sales emails├── Ad copy├── Landing pages├── Headlines & CTAs├── Cold outreach├── Email sequences└── Value propositions **Trigger phrases:** "write email copy", "headline variations","landing page copy", "ad copy", "cold email", "sales page" **Method:** Principles + pattern libraries from 12 legendary copywriters(not rigid templates—situational choices)Claude recognizes the task
The skill identifies the type of copy—thought leadership, persuasive, outbound—and loads the right approach.
## The Productive Tensions Great copy resolves these debates: | Tension | Master A | Master B | Resolution ||---------|----------|----------|------------|| **Proof vs Humanity** | Hopkins: Add numbers | Halbert: Sound human | Numbers *in human voice* || **Their Words vs Promise** | Wiebe: Customer language | Caples: Transformation | Their words, YOUR transformation || **Facts vs Remarkable** | Ogilvy: Pack in facts | Godin: One remarkable claim | Remarkable *facts* win || **Simple vs Targeted** | Gerhardt: Mom-explainable | Kennedy: Hyper-targeted | Simple *for that buyer* || **Position vs Awareness** | Dunford: vs alternatives | Schwartz: Match awareness | Alternatives *they're considering* | **The insight:** When masters conflict, find the synthesis.Borrow from the best
Skills turn expertise into code—principles, patterns, and tensions that inject directly into the model. Claude doesn't guess. It knows what the masters know.
## Loading Training Examples Scanning corpus: 04_projects/gtm-copy-training/corpus/ Found relevant examples:├── cold-email/pvp-gold-standard.md├── headlines/caples-transformation.md├── landing-pages/conversion-optimized.md└── card-copy/proven-hook-patterns.md Loaded 4 curated examples matching "persuasive headline" Pattern detected: Short punch hook → Medium outcome → Short proofTarget: 35-40 words (vs. 50-70 dense blocks)Trained on what actually works
I scraped 4,500 pieces of sales copy and built agents to score them on persuasiveness. The skill loads the winners. You get patterns that actually convert.
## Headline Under Review "Knowledge graph organizes your content" --- **Ogilvy (Researcher):** "Organizes" does zero work. How much? How fast?→ "2,847 files indexed. Found in 3 seconds." **Hopkins (Scientist):** No numbers. No proof. Could be any tool.→ "Search returns results in 0.3 seconds." **Halbert (Storyteller):** Sounds like a product manager wrote it.→ "That doc from 3 months ago? Found." **Caples (Tester):** Would I click this? No transformation shown.→ "From 'I know I wrote that' to 'here it is.'" **Wiebe (Voice Miner):** "Knowledge graph" is vendor jargon.→ "Find anything you've ever written." **Gerhardt (Humanizer):** Mom has no idea what a knowledge graph is.→ "Your files, connected and searchable. Finally." --- **Synthesis:** Transformation + specificity + their language→ "From 'where did I put that?' to 'here it is.' 3 seconds."Faster, not hands-off
Skills have opinions. Sometimes strong ones. That's good—it gives you options. But at the end of the day, you're still in control of how you use them.
The 42 Production Skills
Content & Editorial Skills
10/produce-content
One idea → blog, newsletter, social, images
/edit-content
Rough draft → content you'd actually publish
/final-content-reviewer
Multi-agent quality gate before publishing
/persuasive-copywriting
12 masters framework applied to any copy
/cold-outbound-writer
Research-backed cold emails that get replies
/schedule-content-topic
Know what to write next without deciding
/generate-image
Content → branded image prompts
/generate-video
Spec → video production brief
/seo-audit
Technical SEO, on-page, content quality, and site-type checklists
/email-sequence
Multi-email sequences for outbound, nurture, and onboarding
Design Skills
5/design-ui
Website designs with Tailwind CSS specs
/frontend-coder
React/Next.js implementation from designs
/physical-doc-designer
Document design for print and PDF
/data-story-visualizer
Visual narratives for carousels and slides
/ux-expert
User flows, wireframes, and IA
Intelligence & Research Skills
9/icp-builder
9-agent pipeline → ICP model, personas, scoring, and discovery
/research-prospect
Company → full ABM research with ICP scoring
/research-individual-prospect
Person → dossier with MEDDPICC, hooks, and timing
/competitive-intel
Competitor data → ready-to-use battlecards
/call-synthesis
Transcript → MEDDPICC extraction
/meeting-prep
Meeting context → attendee dossiers adapted by type
/librarian
Query → relevant context surfaced instantly
/knowledge-synthesizer
10-100 documents → structured synthesis
/user-context-gathering
Vague requirements → clear specifications
Planning & Strategy Skills
6/deep-planning
Complex task → step-by-step execution plan
/ralph-execute
Long-running tasks → survive context loss
/software-feature-planner
Feature → architecture and task specs
/spec-first-dev
Multi-agent projects with context preserved
/test-first-code-workflow
Test-first workflow for confident code
/agent-pattern-documenter
Completion → patterns extracted for next time
Validation & Quality Skills
4/skeptical-buyer
Content critiqued from buyer perspective
/prompt-creator
Need → production-quality prompt design
/legal-reviewer
Contract → risk-scored findings
/review-content
Draft → 4-agent quality review before publishing
Utility & Meta Skills
8/agent-memory
Every skill learns from your corrections
/memory-query
Ask what Claude has learned about anything
/push-pr
Save and publish your work with one command. No git knowledge required.
/excel-work
Excel file creation, editing, and analysis
/powerpoint-template-builder
Brand-compliant presentations from client templates
/onboarding
Guided setup for new users
/new-skill-creator
Build your own skills from patterns
/weekly-repo-audit
Automated repository health checks
─── AND FOR YOUR LIFE ───
The same system that runs your content and code
now runs your calendar, kitchen, and budget.
39 skills. Your methods. Production-ready.
Load your context. Apply your standards. Generate ready-to-use outputs.
"This is everything I use. Daily. For 2+ years of production work." —Victor
Get STEEPWORKS