define-customer-profile

**User request**: $ARGUMENTS # Customer Profile Skill Create the foundational CUSTOMER.md document through iterative refinement. This document precisely defines WHO your customer is, their problems, behaviors, and what they value. **Everything else in the product derives from this document.** > **This is the most important document your product will ever have.** Product decisions, feature prioritization, and even brand guidelines all flow from understanding the customer. Get this right first. ## Overview This skill supports both **creating a new customer doc** and **refining an existing one**. Users come back to adjust their ICP as they learn more about their market or pivot their product. **Loop**: Check existing → Discover → Research → Draft → Refine → Repeat until complete This skill guides you through: 0. **Check Existing** - Look for existing CUSTOMER.md; let user choose to refine or start fresh 1. **Discovery** - Clarifying questions about your product, market, and current understanding (keeps asking until confident) 2. **Research Phase** - Launch parallel agents to research ICP data, market, competitors (optional) 3. **Initial Draft** - Generate first CUSTOMER.md based on inputs + research 4. **Refinement & Completion** - Review sections, validate assumptions, update doc, finalize **Returning users** can skip to Phase 4 to refine specific sections. ### Where the Value Comes From The discovery questions give you a solid starting point, but **the real magic happens in Phase 2 (Research) and Phase 4 (Refinement)**: - Research agents dig into real market data, competitor positioning, and ICP characteristics - You validate whether the ICP resonates with your actual experience - Each cycle sharpens the doc until it truly captures YOUR ideal customer **Discovery log**: `/tmp/customer-discovery-{YYYYMMDD-HHMMSS}.md` - external memory updated after each step. ## Workflow ### Initial Setup (create todos immediately) **Create todo list** - areas to discover, not steps. List expands as user answers reveal new areas. **Starter todos**: ``` - [ ] Check for existing CUSTOMER.md; done when existing file read or absence confirmed - [ ] Discovery questions→log; done when user answers captured - [ ] Research→log (if requested); done when research findings logged - [ ] (expand: areas as discovery reveals) - [ ] Generate initial draft; done when draft created - [ ] Refinement cycles→log; done when user approves direction - [ ] Refresh: read full discovery log - [ ] Finalize document; done when CUSTOMER.md written ``` **Create discovery log** at `/tmp/customer-discovery-{YYYYMMDD-HHMMSS}.md`: ```markdown # Discovery Log: Customer Profile Started: {timestamp} ## Discovery Answers (populated incrementally) ## Research Findings (populated incrementally) ## Decisions Made (populated incrementally) ## Refinement Notes (populated incrementally) ``` ### Phase 0: Check for Existing Document **Mark "Check for existing CUSTOMER.md" todo `in_progress`.** Before starting discovery, check if the user already has a CUSTOMER.md: 1. **Search for existing doc**: Use Glob to search for `**/CUSTOMER.md` in the current directory and common locations 2. **If found**: Read it and ask the user what they want to do ``` header: "Existing Customer Doc Found" question: "I found an existing CUSTOMER.md. What would you like to do?" options: - "Refine it - update specific sections based on new learnings" - "Start fresh - create a new customer profile from scratch" - "Review it - just read through what's there" ``` **If "Refine it"**: Ask which sections need updating, then jump to Phase 4 with targeted questions. **If "Start fresh"**: Proceed to Phase 1 (Discovery) - will overwrite the existing doc. **If "Review it"**: Read and display the doc, then ask what they want to do next. 3. **If NOT found**: Proceed directly to Phase 1 (Discovery) **Mark "Check for existing CUSTOMER.md" todo `completed`.** ### Phase 1: Discovery **Mark "Discovery questions" todo `in_progress`.** Use AskUserQuestion tool with multi-choice options for EVERY question. **Put the recommended option FIRST** with "(Recommended)" suffix to reduce cognitive load. **After EACH question**, append to discovery log: ```markdown ### Q{N}: {question topic} **Answer**: {user's answer} **Impact**: {what this reveals about ICP} **New areas**: {any new todos to add} ``` **Recommendation Strategy**: Since there's no prior document, recommendations are based on: - Most common patterns for solo devs/indie hackers - Product type (once selected, informs subsequent recommendations) - Sensible defaults that apply to most cases **Question 1: Product Stage** ``` header: "Product Stage" question: "Where is your product in its lifecycle?" options: - "Early stage - have some users, finding PMF (Recommended)" - "Pre-launch - still building, no customers yet" - "Growth stage - PMF found, scaling acquisition" - "Mature - established product, optimizing" ``` Recommendation: "Early stage" is most common for solo devs using this tool. **Question 2: Product Type** ``` header: "Product Type" question: "What type of product are you building?" options: - "B2B SaaS - software for businesses (Recommended)" - "Developer tool - for engineers/developers" - "B2C app - consumer application" - "Marketplace/Platform - connecting buyers and sellers" - "Content/Media - newsletter, course, community" - "Physical product - hardware or tangible goods" ``` Recommendation: "B2B SaaS" is most common. Use selected type to inform later recommendations. **Question 3: Problem Space** (Free text) ``` header: "Problem" question: "What core problem does your product solve?" freeText: true placeholder: "Describe the main pain point (e.g., 'Teams waste hours manually syncing data between tools')" ``` This is essential context that's hard to multiple-choice. Get 1-2 sentences. **Question 4: Current Customer Knowledge** ``` header: "Current Understanding" question: "How well do you know your ideal customer today?" options: - "Somewhat - have some customers, seeing some patterns (Recommended)" - "Very well - I've talked to many, have clear patterns" - "Vaguely - have hypotheses but not validated" - "Not at all - need to figure this out from scratch" ``` Recommendation: "Somewhat" - most users have some signal but need help structuring it. **Question 5: Primary Value Proposition** ``` header: "Value Prop" question: "What's the PRIMARY value you deliver?" options: - "Save time - automation, efficiency (Recommended)" - "Save money - cost reduction, better ROI" - "Make money - revenue generation, growth" - "Reduce risk - security, compliance, reliability" - "Improve quality - better outcomes, fewer errors" - "Enable capability - do something previously impossible" - "Provide enjoyment - entertainment, satisfaction" ``` Recommendation: "Save time" is the most common value prop for SaaS products. **Question 6: Purchase Decision** Recommendation based on product type: - B2B SaaS → "Team lead" or "Individual user" - Developer tool → "Individual user" - B2C → "Individual user" - Enterprise → "Multiple stakeholders" ``` header: "Buyer Type" question: "Who makes the purchase decision?" options: - "Individual user - they buy for themselves (Recommended for B2C/dev tools)" - "Team lead - buys for their team (Recommended for B2B SaaS)" - "Executive/C-suite - strategic purchase" - "Procurement/IT - goes through formal process" - "Multiple stakeholders - committee decision" - "No purchase - free product, other monetization" ``` **Question 7: Known Customer Characteristics** ``` header: "Customer Traits" question: "What do you know about your best customers? (Select all that apply)" options: - "Pain intensity (desperate vs nice-to-have) (Recommended)" - "Specific job title or role (Recommended)" - "Behavioral patterns (power users, specific workflows)" - "Company size or type" - "Technical skill level" - "Specific industry or vertical" - "Geographic location" multiSelect: true ``` Recommendation: "Pain intensity" and "Job title/role" are the most actionable traits for targeting. **Question 7b: Trait Details (Follow-up)** For EACH trait selected in Q7, ask a targeted follow-up to capture specifics: | If Selected | Follow-up Question | |-------------|-------------------| | Job title/role | "What specific titles? (e.g., 'Engineering Manager', 'Head of Product')" | | Company size | "What size range? (e.g., '10-50 employees', 'Series A-B startups')" | | Technical skill | "What skill level? (e.g., 'Can write code', 'Uses no-code tools')" | | Industry | "Which industries? (e.g., 'Fintech', 'Healthcare SaaS')" | | Geographic | "Which regions? (e.g., 'US-based', 'English-speaking markets')" | | Behavioral | "What behaviors? (e.g., 'Uses Slack daily', 'Already has a workflow')" | | Pain intensity | "How desperate? (e.g., 'Hair on fire', 'Nice efficiency gain')" | Use free-text for these follow-ups - the specifics matter and are hard to predict. **Question 8: What They're NOT** Recommendation based on product type: - B2B SaaS (small team) → Recommend "Enterprise" as anti-persona - Developer tool → Recommend "Beginners" as anti-persona - B2C consumer → Recommend "Enterprise" as anti-persona ``` header: "Anti-Personas" question: "Who is explicitly NOT your customer? (Select all that apply)" options: - "Enterprise (too slow, complex sales) (Recommended for solo devs)" - "Price-sensitive buyers (race to bottom) (Recommended)" - "Beginners (need too much hand-holding)" - "SMB (can't afford, high churn)" - "Experts (don't need the product)" - "Specific industry (bad fit)" - "Not sure yet - need to figure this out" multiSelect: true ``` Recommendation: Most solo devs should avoid "Enterprise" (sales cycle too long) and "Price-sensitive" (race to bottom). **Question 8b: Anti-Persona Details (Follow-up)** For EACH anti-persona selected in Q8 (except "Not sure yet"), ask why: | If Selected | Follow-up Question | |-------------|-------------------| | Enterprise | "Why avoid enterprise? (e.g., 'Sales cycle too long', '6+ month deals kill us')" | | SMB | "Why avoid SMB? (e.g., 'Churn too high', 'Can't afford $X/mo')" | | Beginners | "Why avoid beginners? (e.g., 'Support burden', 'Don't understand the value')" | | Experts | "Why avoid experts? (e.g., 'Build their own', 'Our solution is too basic')" | | Price-sensitive | "Why avoid price-sensitive? (e.g., 'Race to bottom', 'High churn')" | | Specific industry | "Which industries and why? (e.g., 'Healthcare - compliance nightmare')" | Use free-text - understanding the WHY behind anti-personas is critical. **Question 9: Research Needs** Recommendation based on customer knowledge (Q4): - "Not at all" or "Vaguely" → Recommend "Full research" - "Somewhat" → Recommend "Light research" - "Very well" → Recommend "No research" ``` header: "Research" question: "Do you want me to research your market and competitors to inform the ICP?" options: - "Light research - just validate my assumptions (Recommended)" - "Yes, full research - ICP patterns, competitors, market data (takes longer)" - "No research - I have enough context, just help me structure it" ``` Default recommendation: "Light research" balances speed with validation. **Question 10: Additional Context** (Free text) ``` header: "Context" question: "Anything else I should know?" freeText: true placeholder: "Product name, URL, existing customers, specific hypotheses, competitors you know about..." ``` **Question 11: Current State & Triggers** (Free text) ``` header: "Current State" question: "What does your ICP do TODAY to solve this problem (before your product)?" freeText: true placeholder: "e.g., 'They use spreadsheets and manually update them weekly' or 'They don't solve it at all'" ``` ``` header: "Triggers" question: "What triggers them to actively seek a solution?" freeText: true placeholder: "e.g., 'When they miss a deadline due to the problem' or 'When a new team member joins and asks why they do it this way'" ``` **Question 12+: Gap-Filling Questions** After the core questions, assess whether you have enough clarity to proceed. If ANY of these are unclear, ask follow-up questions: | Gap | Follow-up Needed | |-----|-----------------| | Unclear ICP boundaries | "Where exactly is the line between good and bad customers?" | | Vague pain points | "Can you give me a specific example of when this pain happens?" | | Unknown current state | "Walk me through what they do today, step by step" | | Unclear triggers | "What's the moment when they realize they need to solve this?" | | Missing context | "What else should I know about their world?" | **Keep asking until you are highly confident** about: 1. Who the ICP is (specific, not vague) 2. Who they are NOT (clear anti-personas) 3. What problem they have (specific pain) 4. What they do today (current state) 5. What triggers them to seek a solution 6. How research should be focused (if requested) **Todo Expansion Triggers** (add todos when user reveals): | User Answer Reveals | Add Todo For | |---------------------|--------------| | Multiple customer segments | Each segment's characteristics | | Complex buying process | Purchase journey mapping | | Industry-specific needs | Industry research | | Unclear anti-personas | Anti-persona validation | | New pain points | Pain point prioritization | Only proceed to Phase 2/3 when gaps are filled. **Mark "Discovery questions" todo `completed`.** ### Phase 2: Research Phase (If Requested) **If user requested research, mark "Research" todo `in_progress`.** If user requested research, launch 2-3 parallel opus agents to gather data. **Skip this phase if user said "No research"**. **IMPORTANT**: Launch these agents IN PARALLEL (single message with multiple agent invocations): **Agent 1: ICP Pattern Researcher** ``` Launch Task agent (subagent_type: general-purpose, model: opus) with prompt: "Research ideal customer profile patterns for [product type] in the [problem space]. Find: 1. Common job titles and roles of buyers 2. Company characteristics (size, stage, industry) 3. Behavioral indicators of high-intent buyers 4. Pain points that drive purchase decisions 5. Typical objections and concerns