7 Claude Code skills that run your go-to-market motion from the terminal.
Each skill encodes a GTM methodology — not a prompt, a repeatable system that gets sharper as you use it. Install them, trigger them by name, and run your GTM from Claude Code.
| # | Skill | What it does |
|---|---|---|
| 01 | ICP Sharpener | Structured interview that produces a scored ICP with segments, buying signals, deal economics, and disqualifiers |
| 02 | LinkedIn Voice Builder | Analyzes your writing + runs a voice interview → produces a 3-mode voice doc (Teaching, Storytelling, Provoking) |
| 03 | LinkedIn Daily Debrief | 5-minute daily capture with rotating question sets → content seeds with urgency tags and auto-review |
| 04 | LinkedIn Idea Sourcer | Searches Reddit, HN, and industry discourse → 5 ideas with specific post outlines and gap detection |
| 05 | LinkedIn Post Repurposer | Takes any content, deconstructs it visibly, finds your angle, rewrites in your voice with format options |
| 06 | Meeting Prep Brief | One-page brief adapted by meeting type (discovery/demo/negotiation/renewal) with competitive intel |
| 07 | Outreach Packager | Researches prospects, finds signals + warm paths, generates a 3-touch sequence with A/B variants |
Clone this repo into your project:
mkdir -p .claude/skills
git clone https://github.com/Othmane-Khadri/gtm-agent-playbook.git temp-playbook
cp -r temp-playbook/.claude/skills/gtm-playbook .claude/skills/
rm -rf temp-playbookOr copy the .claude/skills/gtm-playbook/ directory manually into your project's .claude/skills/.
Claude Code auto-discovers skills — no configuration needed.
Trigger: Say sharpen my ICP, define my ICP, or who should I sell to
What happens:
- Asks how many paying customers you have
- 5+ customers → Evidence Mode (7 questions grounded in real conversion data)
- <5 customers → Hypothesis Mode (5 questions + web research to validate assumptions)
- Produces a scored ICP document with one-line ICP statement, deal economics, segments, a 4-axis scoring rubric (Fit/Timing/Access/Intent), buying signals, disqualifiers, and competitive landscape
- Saves to
docs/icp.md— referenced automatically by Idea Sourcer and Outreach Packager
Revisit: Say recalibrate my ICP after 3+ months. It re-interviews against actual conversion data and re-weights the scoring rubric.
Trigger: Say build my LinkedIn voice or define my writing style
What happens:
- Collects 3-5 writing samples (including your best-performing post)
- Analyzes across 7 dimensions (sentence rhythm, hooks, structure, tone, formatting)
- Runs a 5-question voice discovery interview focused on audience, influences, and feel
- Produces a voice doc with 3 writing modes (Teaching, Storytelling, Provoking), tone spectrum, hook patterns, word substitutions, and LinkedIn-specific platform rules
- Saves to
docs/voice.md— referenced automatically by Daily Debrief, Idea Sourcer, and Post Repurposer
Check consistency: Say check my voice after 5+ posts. It scores each post as a percentage match against your voice doc and surfaces drift patterns.
Trigger: Say daily debrief or debrief me
What happens:
- Asks 3 questions from a rotating set (3 different sets cycle by day to prevent fatigue)
- Extracts 2-3 content seeds, each tagged with angle, audience, urgency (🔴 48h / 🟡 this week / 🟢 evergreen), and topic tags
- Appends to
docs/content-seeds.md(accumulates over time, never overwrites) - At 7+ seeds, automatically surfaces top 3 to write first
Review seeds: Say review my seeds to prioritize, cluster by theme, and prune stale ideas.
Best practice: Run this daily for a week. You'll bank 14-21 seeds and start seeing your signature topics emerge.
Trigger: Say source ideas, find content ideas, or what should I write about
What happens:
- Loads your ICP and voice doc for context (if they exist)
- Searches Reddit, Hacker News, Twitter/X, and industry sources for what's trending in your niche
- Cross-references against your content seeds to avoid duplicates
- Produces 5 idea cards, each with: trending topic, your unique angle, a specific hook, a 4-line post outline, and resonance rating
- At least 2 of 5 are contrarian or surprising angles
Track performance: Say which ideas worked after publishing. It logs results and weights future ideas toward angles that performed.
Requires: Web search (Claude Pro/Team plan or configured MCP)
Trigger: Share a URL or paste content and say repurpose this or make this mine
What happens:
- Loads your voice doc (or asks 3 quick questions if you don't have one — it won't block you)
- Fetches and deconstructs the source content — shows you the breakdown so you learn how to think about content
- Asks 2-3 questions to find YOUR angle on the idea
- Lets you choose format: text post, carousel, or thread
- Rewrites entirely in your voice — different hook, structure, and framing. Never references the original.
- Shows what it kept, what it changed, and which voice patterns it applied
Track performance: Say how did the repurposed post do to log results and surface which angles and formats perform best.
Trigger: Say prep me for [company] or meeting prep
What happens:
- Asks for company, contacts, and meeting type (discovery / demo / negotiation / renewal / partnership)
- Runs 6-8 web searches for company context, recent signals, competitive landscape, and contact profiles
- Generates a brief adapted to the meeting type:
- Discovery → qualification questions (BANT), talking points about their problems
- Demo → features mapped to their pain points
- Negotiation → leverage points, value-based talking points
- Renewal → ROI recap, expansion opportunities
- Adds stakeholder map for multi-person calls
- Saves to
docs/briefs/[company]-[date].md
Post-call: Say debrief the call to capture what worked, what surprised you, and next steps. Each debrief appends to the same file — by meeting 3, you have a full deal history.
Requires: Web search (Claude Pro/Team plan or configured MCP)
Trigger: Say package outreach for [names] or share a prospect list
What happens:
- Loads your ICP for fit scoring (if it exists)
- Researches each prospect (max 5 per batch): company context, personal signals, and warm paths (shared communities, events, connections)
- Generates a 3-touch sequence per prospect:
- Day 0: Signal-based opener (1-2 sentences for DM, 3-5 for email) + A/B variant
- Day 3: Value drop — share a resource, no ask
- Day 7: Direct ask with an easy out
- Recommends the best channel (LinkedIn DM vs email) based on prospect's online activity
- If no public signal exists, uses an industry-angle approach instead of faking it
Track replies: Say track replies to log which signals, channels, and touch numbers get responses. After 3+ batches, it recommends ICP refinements based on actual reply data.
Requires: Web search (Claude Pro/Team plan or configured MCP)
ICP Sharpener ──→ docs/icp.md ──→ Idea Sourcer (audience context)
→ Outreach Packager (fit scoring)
Voice Builder ──→ docs/voice.md ──→ Daily Debrief (angle matching)
→ Post Repurposer (voice modes)
→ Idea Sourcer (hook patterns)
Daily Debrief ──→ docs/content-seeds.md ←── Idea Sourcer (overlap check)
Start with Skills 01 and 02. Everything downstream reads their output automatically.
- Claude Code installed
- Claude Pro or Team plan recommended — Skills 04, 05, 06, and 07 use web search
- A project directory where you work (the skills save output to
docs/in your project root)
- Run the Daily Debrief for 5 days. You'll bank 10-15 content seeds with urgency tags.
- Outreach Packager works best with 3-5 prospects at a time. Research quality drops with large batches.
- Meeting Prep works for any meeting — investor calls, partnerships, podcast interviews. Just pick the right meeting type.
- Skills improve with use. Voice Builder learns from corrections. Content seeds accumulate. ICP gets sharper each revisit. Outreach sequences optimize based on reply data.
- Everything saves to
docs/. Your ICP, voice, seeds, briefs, and logs all live in one place and build on each other.
Built by Othmane Khadri. GTM engineering for B2B SaaS — AI agents, content systems, and outreach infrastructure.