Setup

Ask these questions ONE AT A TIME. Wait for each answer before asking the next. Do not ask all at once.

Present each question as a short, friendly prompt. After each answer, acknowledge it briefly (one line) then ask the next.

Q1 — Who are you?

"First — are you setting this up for yourself (individual creator) or for a company/brand?"

• Individual → proceed with "I/me" framing throughout
• Company → use "we/our" framing, ask for company name + industry instead of personal details

Q2 — Name and role

"What's your name and title? (e.g., Jane Smith, ML Engineer)"

Capture: name, title

Q3 — Company and location

"Where do you work and where are you based? (e.g., Stripe, Berlin, Germany)"

Capture: company, location

Q4 — Your specialization

"What's your technical or professional specialization? In 1-2 sentences — what do you actually work on day to day?"

Example answers:

• "I build NLP systems for legal document processing"
• "I'm a DevOps engineer focused on Kubernetes and platform engineering"
• "We make AI tools for e-commerce product teams"

Capture: specialization (rewrite into a clean 1-liner)

Q5 — Your positioning (the most important question)

"What makes you different from everyone else with your job title? Think: what have you SHIPPED or BUILT that most people in your role haven't? Give me one specific thing."

Example answers:

• "I deployed a clinical speech recognition system that keeps all data on-premise in hospitals"
• "I've processed 2M+ legal documents in production using LLMs, not just built demos"
• "We're the only AI team building for Urdu-language healthcare in Pakistan"

Capture: positioning_statement

Rewrite it into a strong positioning statement in first person:

"I am one of the few [title] who has [specific achievement] — not just [what most people do]."

Show the rewritten version to the user and ask: "Does this sound like you? Say yes or tweak it."

Q6 — Proof points

"Give me 2-3 specific projects or results with real numbers. Format: [Project name] — [what it does] — [result/metric]"

Example:

• "Chest X-ray detection — deployed in Pakistani hospitals — ~93% accuracy"
• "Legal RAG pipeline — processes 10K docs/day — 40% faster than previous system"

Capture: proof_points[] (list of 2-3 items)

Q7 — Content pillars

"What 3-4 topics do you want to post about? These become your content pillars. List them — e.g., 'Healthcare AI, Python tutorials, Career in tech, Industry news'"

For each pillar, automatically assign a percentage split:

• Pillar 1 (primary expertise): 35%
• Pillar 2 (technical depth): 25%
• Pillar 3 (personal/career): 20%
• Pillar 4 (commentary/news): 20%
• If 5 pillars: redistribute as 30/20/15/15/10 (user can adjust later)

Capture: pillars[] (list of pillar names)

Q8 — Target audience

"Who are you trying to reach? List 2-3 types of people — e.g., 'recruiters in healthcare AI, other ML engineers, startup founders who need AI help'"

Capture: audience[]

Q9 — Timezone and posting schedule

"What timezone are you in, and what time works best for you to post? LinkedIn's best window is usually 8-10 AM your local time."

Example: "PKT, 8-10 AM" or "EST, 9 AM"

Capture: timezone, post_time

Q10 — Brand color (optional)

"One last thing — do you have a brand accent color you prefer? (e.g., #00D4AA teal, #FF6B35 orange, #6C63FF purple). Or just say 'default' to keep teal."

Capture: accent_color (default: #00D4AA)

Step 2: Confirm Before Writing

After collecting all 10 answers, show a brief summary:

Here's what I'll set up for you:

  Name:         [name]
  Title:        [title]
  Company:      [company], [location]
  Positioning:  "[positioning_statement]"
  Pillars:      [pillar 1] (35%) · [pillar 2] (25%) · [pillar 3] (20%) · [pillar 4] (20%)
  Audience:     [audience summary]
  Post time:    [post_time] [timezone]
  Accent color: [accent_color]

Writing 6 config files... (type 'wait' to change anything first)

Pause for 3 seconds conceptually — if user says nothing or says "go", proceed.

Step 3: Write All 6 Files

Write these files using the collected answers. Use the template files as structural reference but populate them fully — no [PLACEHOLDER] strings should remain in the output.

File 1: context/profile.md

# Profile — [name]

## Identity
- **Name:** [name]
- **Title:** [title]
- **Company:** [company]
- **Location:** [location]
- **Specializations:** [specialization — split into 2-3 bullet tags]
- **Mission:** [1-line mission derived from positioning]

## Unique Positioning Statement
> "[positioning_statement]"

This is the moat. Every piece of content should reinforce this positioning, directly or indirectly.

## Key Projects & Technical Proof Points
[For each proof point in proof_points[]:
- **[Project Name]** — [description and result]
]

## Voice & Tone
- **Conversational and direct** — write like talking to a smart friend
- **Technical but accessible** — deep expertise, no jargon for jargon's sake
- **Specific over vague** — real numbers, real systems, real challenges
- **Honest** — share failures and what they taught, not just wins
- **Confident but not preachy** — share what's working, acknowledge what isn't
- **Simple English** — short sentences, short paragraphs
- **No corporate speak** — banned: synergy, leverage, ecosystem, excited to announce, game-changer

## Social Profiles
- LinkedIn: [ADD YOUR LINKEDIN URL]
- GitHub: [ADD YOUR GITHUB URL]

File 2: config/brand.json

Populate with:

• brand_name: "[name] — [title]"
• tagline: derive a 1-line tagline from the positioning statement
• colors.accent: [accent_color]
• linkedin_profile.headline: derive from title + specializations
• linkedin_profile.cta: "DM me about [specialization]"
• carousel.cta_slide_text: "Follow for more [pillar 1] content"
• Keep all structural fields (fonts, dimensions, style) unchanged from the template

File 3: config/content-pillars.json

For each pillar the user gave:

• Generate 4-6 realistic topic ideas based on the user's specialization and proof points
• Assign appropriate hashtags (5-6 per pillar, relevant to the topic)
• Write a goal and tone for each pillar
• Map posting schedule to Mon/Wed/Fri with the user's timezone and time
• Keep all structural fields (format_priority, hashtag_rules) unchanged

File 4: context/strategy.md

Write a full strategy doc using the pillar structure. For topics under each pillar, generate specific ideas grounded in the user's actual proof points — not generic examples.

File 5: config/audience-personas.json

For each audience the user listed, generate a full persona with:

• Realistic description (2 sentences)
• 2 pain_points relevant to their role
• what_they_want_from_you — specific to the user's positioning
• content_that_resonates — which post types work for them

File 6: research/hooks-bank.md

Generate 5-7 starter hooks per pillar, grounded in:

• The user's actual proof points (use their real numbers)
• Their positioning statement
• The pattern: curiosity gap, bold claim, specific number, or provocative question

Step 4: Confirm and Next Steps

After writing all files, say:

✓ Setup complete. 6 files written:
  • context/profile.md
  • context/strategy.md
  • config/brand.json
  • config/content-pillars.json
  • config/audience-personas.json
  • research/hooks-bank.md

You're ready to start. Run /prime to initialize your session,
then /research to scan this week's trends.

Important Rules

• Never leave a [PLACEHOLDER] string in any output file — fully populate everything from the answers
• If the user gives vague answers (e.g., "I do AI stuff"), ask one follow-up to get something specific before moving on
• Keep questions short — this is a conversation, not a form
• The proof points are critical — if the user doesn't give specific numbers, prompt once: "Can you give me a real metric? Even approximate — like ~90% accuracy or 3x faster?"
• Do not ask more than 10 questions total (consolidate if needed)