Your AI content system, built in an afternoon.
A step-by-step setup that turns any coaching call, voice note, or raw material into polished content on your schedule. Five steps. Each with a copy-paste prompt. You never have to wonder what to do next.
Start with Step 1 →Five steps from raw material to finished posts.
This system takes the content you produce every day - coaching calls, voice notes, client stories, workshop recordings - and turns it into a pipeline of ready-to-post content that sounds exactly like you.
This is the intake. Pull together everything you already have in one fresh conversation. The AI reviews it, tells you what it understands, and surfaces any missing pieces.
- Your three brand documents (Magnetic Messaging Map, Ideal Client DNA, Authentic Voice Guide)
- Your top 10 to 30 posts - whichever you're most proud of, or that got the strongest response
- Any raw material you have: coaching call transcripts, voice notes, workshop recordings, client DMs, testimonials, Fathom exports
How to do it
- Open a fresh conversation in Claude (or your chosen AI)
- Paste or upload everything from the list above into that one conversation
- Copy the prompt below and paste it underneath your uploaded material
- Answer any gap questions the AI asks, one at a time
I've uploaded everything I have above - my three brand documents (Magnetic Messaging Map, Ideal Client DNA, Authentic Voice Guide), my top-performing posts, and any raw material I have (call transcripts, voice notes, testimonials, coaching notes, workshop recordings). Your job is to read all of it carefully and give me three things, in this order. --- 1. INVENTORY - What You Now Know About Me Produce a short, specific summary covering: - My business identity and offer (one paragraph, drawn directly from my Magnetic Messaging Map) - My ideal client (one paragraph, drawn from my Ideal Client DNA - use their actual language, not generic descriptions) - My voice patterns (top 5 phrases I repeat, my typical opening style, my typical closing style) - My content patterns (the 3 to 5 formats I produce most often, which platforms, what a typical piece looks like) - My proof points and credentials (real numbers, client outcomes, milestones visible in my material) Be specific. If you can quote a real line from what I gave you, quote it. Generic language here means you didn't read closely enough. --- 2. GAPS - What's Missing List any critical piece of material that would sharpen the system but isn't here yet. Common gaps: - No call transcripts or Fathom exports, so no raw material to mine - Voice samples are all one format, so I can't see cross-format consistency - No testimonials in the client's own language, so I can't mirror how they speak - No specific numbers or proof points, so posts will feel vague Rank gaps by impact. The #1 gap is the one that most limits what the system can produce. --- 3. READINESS CHECK - On a Scale of 1 to 10 How ready am I to run content mining and drafting on autopilot? If the score is below 8, tell me the ONE thing I should fix first before moving to Step 2. --- After you've delivered all three sections, ask me about the most important gap first. One question at a time. Wait for my answer before moving to the next. When I've either filled the gaps or explicitly said "skip for now," tell me I'm ready for Step 2.
A skill is a reusable recipe for producing one type of content in your voice. Instead of using someone else's template, we build yours - one for each format you care about. You'll end up with a small library of skills you can install into Claude and trigger by name.
Why this step is non-negotiable
Yes, you could skip straight to mining and drafting. Technically, Claude can draft anything without a custom skill. But without skills tuned to YOUR voice, Step 4 and Step 5 produce generic AI output - the exact thing your audience scrolls past.
Custom skills are the leverage layer. They're what turn this system from "a fancy ChatGPT wrapper" into "a content machine that produces work your audience thinks you sat down and wrote yourself." Build at least your top 2 formats before moving on.
Run this prompt ONCE PER SKILL you want to build. It interviews you about a specific content format, studies your top posts, and produces a complete .skill file you can install. Start with the format you produce most often. Then repeat for each additional format.
How to do it
- Open a fresh conversation (ideally the same one from Step 1, so the AI already knows your material)
- Copy the prompt below and paste it
- Answer the content-pattern questions one at a time - be specific, pull from your actual posts
- The AI will produce a complete .skill file. Save it and follow the install steps it gives you.
- Start a fresh conversation and repeat for the next format
You are my Custom Skill Designer. Your job is to create a complete, installable Claude Skill that writes ONE specific content format in my voice, for my audience, using my patterns. The final output is a .skill file I can install into Claude and trigger by name.
You already have my three brand documents and my top posts from Step 1. Use them as your source of truth. Do not fabricate voice patterns, phrases, or examples - pull everything from my real material.
---
PHASE 1 - PICK THE SKILL
Ask me this exact question:
"Which format do we build a skill for today? Pick one. Examples: Instagram Reels caption, Instagram Story, Facebook Top-of-Funnel post (thought leadership, belief busters, frameworks, personal stories), Facebook Middle-of-Funnel post (client story posts), Email, Carousel, B-roll caption, YouTube hook, or something else specific to you."
Wait for my answer.
---
PHASE 2 - CONTENT PATTERN INTERVIEW
Ask me ONE question at a time. Wait for my answer before moving to the next.
1. "From my uploaded top posts, pull the 3 to 5 that are examples of this format. Show them to me and tell me what you observe about their structure, rhythm, and voice." (You propose first - then I confirm or correct.)
2. "What is the GOAL of this format for me - audience building, lead nurture, direct response, community warm-up, or something else? What outcome tells me a piece of this format worked?"
3. "Who EXACTLY is the reader of this format?" (Draw from my Ideal Client DNA. If my audience for this format is different from my general ideal client, I'll say so.)
4. "Based on my examples, here's the structure I see: [propose hook, middle, close in detail]. Is this right? What am I missing or getting wrong?"
5. "What's the ONE thing I do in this format that nobody else does? The non-obvious pattern, the counterintuitive choice, the specific move that makes this unmistakably mine."
6. "What should NEVER appear in this format? Pull from my Authentic Voice Guide - banned phrases, banned patterns, tones to avoid. Add anything else format-specific."
7. "What's the ideal length? Words, characters, seconds of video, or however this format measures."
8. "What trigger phrases would activate this skill when I type them into Claude? (For example, 'write a FB post', 'draft a reel caption', 'turn this into an email'.)"
9. "What should this skill REFUSE to do? When should it redirect me to a different skill? (For example, if I ask for a client-story post, a TOF skill should redirect me to the MOF skill.)"
---
PHASE 3 - DRAFT THE SKILL FILE
Produce a complete .skill file in this EXACT format. No commentary before or after.
---
name: [skill-name-in-kebab-case]
description: |
[One long paragraph covering: what this skill does, who it's for, when it triggers. Include the exact phrase "MANDATORY TRIGGERS:" followed by 8 to 12 specific trigger phrases. Then include "Do NOT trigger for:" followed by the exclusions from Phase 2 Q9.]
---
# [Skill Title]
## WHO THIS IS FOR
[Reader description from Phase 2 Q3, in 1 to 2 paragraphs]
## CRITICAL RULES
[Numbered list of 6 to 10 hard rules. Pull from banned phrases, non-obvious patterns, and format-specific constraints.]
## STRUCTURE
### 1. HOOK - [Name of the hook style]
[2 to 4 paragraphs explaining exactly how my hooks work in this format, with 2 to 3 real examples pulled directly from my posts.]
### 2. MIDDLE - [Name of the middle section]
[2 to 4 paragraphs on the middle, with real examples.]
### 3. CLOSE - [Name of the close]
[2 to 4 paragraphs on the close, with real examples.]
## VOICE & TONE
**DO:**
[Bulleted list of voice moves that should appear in this format]
**DON'T:**
[Bulleted list of banned moves]
## LENGTH
[Specific target from Phase 2 Q7]
## QUALITY CHECKLIST
[Checklist of 10 to 15 yes/no items to run internally before output. Cover: voice rules, structure rules, length, "does this sound like me?", platform-specific constraints.]
## REFERENCE EXAMPLES
[Include 3 real examples from my uploaded top posts. For each, add a one-paragraph analysis titled "Why this works" that breaks down the hook move, middle move, and close move.]
## OUTPUT FORMAT
Produce ONLY the finished piece - no commentary, no "here's what I wrote." Just the content ready to copy-paste.
Use this structure:
```
{Hook}
{Middle}
{Close}
```
After the piece, include:
`[Length: XXX]`
---
PHASE 4 - INSTALLATION INSTRUCTIONS
After producing the skill file, tell me exactly these three things:
1. HOW TO SAVE THE FILE:
- Create a new folder on your computer named [skill-name-kebab-case]
- Inside that folder, create a file called SKILL.md
- Paste everything between the two --- lines above into SKILL.md
- Zip the folder. Rename the zip's extension from .zip to .skill
2. HOW TO INSTALL IT IN CLAUDE:
- Open Claude Desktop (or Claude.ai Projects)
- Go to your project settings → Skills → Add Skill
- Upload the .skill file
- Confirm it's active
3. HOW TO TEST IT:
- In a new conversation, type one of the trigger phrases followed by raw input material
- Verify the output matches the structure and voice described in the skill
- If it feels off, note what's wrong and rerun this builder prompt to refine the skill
Done. Build one skill at a time. When I'm ready for the next format, I'll start a fresh conversation and paste this prompt again.
Marc's skill bundle - download and study as reference
These are Marc's real production skills, tuned to HIS voice, HIS audience, HIS frameworks. They will NOT sound like you if you install them as-is. Instead of studying them line by line, the fastest path is to download one, paste it into your AI, and say: "Adapt this skill for me. Interview me on my voice, format, and audience, then rewrite every section so it fits my style." Your AI handles the conversion.
How to install a .skill file: open Claude Desktop → Settings → Skills → Add Skill → upload the downloaded file. Reminder: use these as structural reference only. The whole point of Step 2 is to build YOUR versions.
This is the home for every piece of content you'll ever produce. One row = one idea. Fields tell you what format it becomes, which platform it lives on, what status it's in, and how strong it is. Built right, this database is the single source of truth for your entire content operation.
The prompt below interviews you about your formats, platforms, cadence, and workflow, then produces a complete Notion database spec tailored to you. It uses best practices from top content teams and references Marc's schema as a starting structure - but the final database is yours, not his.
How to do it
- Open a fresh conversation with your uploaded material from Step 1
- Copy the prompt below and paste it
- Answer one question at a time - the AI is building YOUR database, so be specific
- Follow the setup instructions the AI gives you to create the database in Notion
You are my Notion Database Architect. Your job is to design a personalised Content Blocks database for me in Notion - the single place where every piece of raw content (coaching calls, voice notes, client stories, frameworks, observations) gets captured, tagged, and assigned to production. Use what you know about me from my three brand documents, my top posts, and my custom skills (from Steps 1 and 2). The database should match MY production reality, not a generic template. --- PHASE 1 - DISCOVERY (one question at a time, wait for answer) 1. "List every content format you produce. Examples: Instagram Reel, Carousel, Facebook TOF Post, Facebook MOF Post, Mentorship Email, YouTube Video, B-roll Caption, Community Post. Include everything, even formats you produce occasionally." 2. "Which platforms do you post to? (Remember: one format can go to multiple platforms. An IG Reel can also be a YouTube Short.)" 3. "What's your weekly production cadence per format? Example: 3 IG Reels, 2 FB Posts, 1 Email, 1 YouTube. Be honest about what you actually ship, not what you aspire to." 4. "Where does your raw content come from? Coaching calls via Fathom? Voice notes? Workshop recordings? Client DMs? Personal journaling? List every source." 5. "Walk me through your status workflow from idea to published. Example: Raw → Scripted → In Production → Scheduled → Published. What are YOUR stages?" 6. "How do you want to grade content quality? Two good options: (a) numeric 1 to 10, or (b) tiered labels like Iconic / Viral / Strong / Good / Filler. Or your own system. Which do you prefer?" 7. "Which script frameworks do you default to? Common ones: Problem-Solution, Myth Buster, Story-Lesson, Contrarian Take, Before-After, List/Tips, Social Proof, Behind the Scenes. Mark which apply, and add any of your own." 8. "How do you categorise content type? Examples: Teaching Moment, Story, Framework, Hot Take, Proof/Result, Objection Handling. Mark which apply, and add your own." 9. "Do you track virality signals separately from overall content rating? (Yes/no. If yes, what signals matter most: specificity, emotional hook, contrarian angle, proof, quotability, CTA clarity, universality?)" --- PHASE 2 - PRODUCE THE DATABASE SPEC Output a complete spec in this exact format: === DATABASE NAME: [Your-Name] Content Blocks PURPOSE: [One sentence on what this database does for me] --- PROPERTIES (fields): 1. Topic / Angle - Title - The one-line angle this block is about. (This is the database's title field.) 2. Block ID - Auto-increment ID - For reference and linking. 3. Status - Select - [My workflow stages from Phase 1 Q5] 4. Format Assignment - Select - [Each format from Phase 1 Q1 as an option, plus "Unassigned"] 5. Best Platform - Multi-select - [Each platform from Phase 1 Q2] 6. Content Type - Select - [From Phase 1 Q8] 7. Script Framework - Select - [From Phase 1 Q7] 8. Content Rating - [Numeric 1-10 OR Select with tier labels, based on Phase 1 Q6] 9. Virality Score - Number 0-7 - [Only include if Phase 1 Q9 is yes] 10. Verbal Hook - Text - First 3 seconds spoken hook for video (if they produce video). 11. On-Screen Text Hook - Text - 6-word-max overlay text for video. 12. Core Insight - Text - One sentence: the belief shift this content creates for the reader. 13. The Actual Story - Long text - Real 2-3 paragraph excerpt from my raw material. NO AI generation, only real words. 14. Source Call / Source Material - Relation or URL - Link back to the coaching call, voice note, or raw material this block came from. 15. Fathom Timestamp (or similar) - URL - Direct link to the moment in the recording where this block appeared. [Only if they use Fathom or a timestamped tool.] 16. Week Assigned - Date - Which week this block is assigned for production. Only include fields that match MY reality from Phase 1. If I don't produce video, skip verbal/visual hooks. If I don't use Fathom, use a generic Source URL field instead. --- SELECT OPTIONS (for every select field, list the options and recommended Notion colors): [Example for Status: Raw (gray), Scripted (blue), In Production (orange), Scheduled (yellow), Published (green)] [... complete for every select field] --- VIEWS (minimum 4): 1. DEFAULT TABLE VIEW - All fields visible, sorted by creation date descending. Use for capture and bulk edits. 2. KANBAN BY STATUS - Grouped by Status. Card shows: Topic/Angle, Format Assignment, Best Platform, Content Rating. Use for daily execution. 3. CALENDAR BY WEEK ASSIGNED - Shows blocks on the week they're scheduled. Use for weekly planning. 4. FILTERED VIEW: READY TO DRAFT - Filter: Status = Scripted AND Format Assignment is not Unassigned. Use this as the queue for Step 5 drafting. 5. [Recommended additional view based on my production reality] - [Purpose] --- AUTOMATION & RITUAL SUGGESTIONS: - Weekly: Open the Calendar view on Monday morning and drag blocks into the current week. - Daily: Open the Kanban view and move one block from Scripted to In Production. - [Other automation suggestions based on my setup] === --- PHASE 3 - SETUP INSTRUCTIONS Give me step-by-step instructions to build this in Notion: 1. HOW TO CREATE THE DATABASE: - Open Notion - Create a new full-page database - Name it [name from spec] - Delete the default properties 2. HOW TO ADD PROPERTIES (in order, easiest to hardest): - Start with the title field (rename "Name" to "Topic / Angle") - Add each property one by one, using exact names and types from the spec - For select fields, add options in the order I listed them (Notion uses this order for sorting) 3. HOW TO CREATE EACH VIEW: - Click "+ Add a view" at the top of the database - Pick the view type (Table, Board, Calendar) - Name the view - Set filters, sorts, and displayed properties exactly as specified 4. WEEKLY RITUAL (one paragraph): - How to capture new blocks (via Step 4 Mining Prompt) - How to pick which blocks to produce this week - How to move blocks through the status workflow --- REFERENCE - MARC'S SCHEMA (starting template, not gospel) For your information, Marc's MI Content Blocks database uses these 16 fields: Topic/Angle, Block ID (auto-increment), Status, Format Assignment, Best Platform, Content Type, Script Framework, Content Rating, Virality Score, Verbal Hook, On-Screen Text Hook, Core Insight, The Actual Story, Source Call (relation), Fathom Timestamp (URL), Week Assigned. He grades content from "10 - Iconic" down to "5 - Filler" and his Script Frameworks are Problem-Solution, Myth Buster, Story-Lesson, Contrarian Take, Before-After, List/Tips, Social Proof, Behind the Scenes. Use this as reference only. Adapt the structure to match MY answers from Phase 1. Do not copy Marc's fields wholesale if they don't match how I actually produce content.
Starter Content Blocks template - duplicate into your workspace
A clean, ready-to-use content blocks database with all 16 fields set up and 3 sample rows to show how the fields work together. Duplicate it into your own Notion, then run Prompt 3 above to customise the fields and options for your specific formats.
How to duplicate: open the link, click the three-dot menu in the top-right → "Duplicate" → select your own workspace as the destination. You'll get a fully editable copy with all fields, views, and sample rows intact. Delete the 3 sample rows once you've added real blocks from your material.
Every week, the mining task runs through whatever raw material you've dropped in (call transcripts, voice notes, workshop recordings) and extracts high-quality content blocks straight into your Notion database. You wake up on Monday to a queue of ready-to-use ideas.
Claude Cowork Recommended
Desktop scheduled tasks run automatically each week. Connect Notion once, drop raw material into a folder, and the blocks land in Notion without you lifting a finger.
ChatGPT or Other AI
No native scheduling. Block a recurring 10-minute slot in your calendar, open the prompt, paste your raw material. The Notion blocks get copied manually.
If you're on Claude Cowork (recommended)
- Connect Notion first. In Claude Desktop → Settings → Connectors → add Notion. Give it access to your Content Blocks database. This is a one-time setup.
- Create a raw-material folder anywhere on your computer (e.g.,
~/Documents/Content-Inbox/). This is where you'll drop transcripts, voice note text, workshop recordings as they come in. - Open Claude Desktop → Schedule sidebar → New Task. Name it "Weekly Content Mining."
- Set frequency: Weekly, Monday 7am (or whenever you plan your week).
- Paste the Mining Prompt below as the task instructions, with your folder path and database name filled in.
- Enable "Keep computer awake during scheduled runs" in task settings, or leave your laptop plugged in and lid-open on Mondays.
- On the first run, approve the file-read and Notion-write permissions when prompted. Choose "Always allow" so future runs go silent.
If you're on ChatGPT or another AI (manual run)
- Pick a day and time each week (Monday morning works well)
- Open a fresh conversation with your brand docs uploaded
- Paste the Mining Prompt below, followed by your raw material for the week
- Copy the output blocks into your Notion database by hand
You are my Content Miner. Your job is to take the raw material I paste or link below (coaching call transcript, voice note, workshop recording, client DM thread, or any long-form content) and extract high-value content blocks straight into my Notion Content Blocks database. --- PROCESS 1. Read the raw material carefully, the whole way through. Do not skim. 2. Identify 3 to 10 distinct content blocks. A block is ONE complete idea that could stand alone as a post, reel, or email. 3. For each block, produce every field from my Notion Content Blocks database schema (see my uploaded material - you know my schema from Step 3). --- FOR EACH BLOCK, OUTPUT THESE FIELDS - Topic / Angle - One line. What this is ABOUT. - Core Insight - One sentence. The belief shift this creates for my reader. - Verbal Hook - First 3 seconds spoken. (Skip if not producing video.) - On-Screen Text Hook - Max 6 words. Overlay text. (Skip if not producing video.) - The Actual Story - A 2-3 paragraph DIRECT excerpt from the raw material. Use my real words. Do not paraphrase. Do not invent. - Content Type - Pick from my schema (Teaching Moment, Story, Framework, Hot Take, Proof/Result, Objection Handling, or whatever my schema defines). - Script Framework - Pick from my schema (Problem-Solution, Myth Buster, Story-Lesson, Contrarian Take, Before-After, List/Tips, Social Proof, Behind the Scenes, or my own). - Best Platform - Pick one or more from my schema. - Format Assignment - Pick ONE format this block is best suited for (or "Unassigned" if multiple fit equally well). - Content Rating - Your honest assessment using my rating system. - Virality Score (0-7) - One point for each signal present: (1) highly specific, (2) emotional hook, (3) contrarian or counter-expected, (4) contains proof or numbers, (5) quotable one-liner, (6) clear CTA, (7) universal relevance. - Status - Default: Raw. - Source - Link back to where this block came from (Fathom URL with timestamp if I gave you one, or a text description). - Week Assigned - Leave blank unless I explicitly tell you which week. --- OUTPUT FORMAT For each block, produce a copy-paste-ready row in this exact structure. Separate blocks with --- dividers. --- BLOCK 1 Topic / Angle: [...] Core Insight: [...] Verbal Hook: [...] On-Screen Text Hook: [...] The Actual Story: [...] [...] [...] Content Type: [...] Script Framework: [...] Best Platform: [...] Format Assignment: [...] Content Rating: [...] Virality Score: [...] /7 Status: Raw Source: [...] --- (If you have Notion write access through an integration, write each block directly into my Content Blocks database and tell me the block numbers and titles you added.) --- HARD RULES - Never fabricate a story, quote, number, name, or detail. If a block would need invented material to work, skip it. - Only extract blocks rated 6 or higher on my rating scale. Quality over quantity - five strong blocks beat fifteen weak ones. - Use my real words from the raw material in "The Actual Story." Always. - If the raw material is too short or contains no extractable blocks, tell me honestly. Don't pad. - If I'm on Claude Cowork and this is running as a scheduled task, check my raw-material folder at [YOUR FOLDER] for any new files since the last run, process them, and write blocks directly into Notion. Ready. Raw material below (or in the folder referenced above).
Cowork task configuration (copy into your scheduled task)
Once Notion is connected and your raw-material folder exists, paste this as the task prompt (replace the two paths with yours):
Read every file in /Users/YOURNAME/Documents/Content-Inbox/ that was added or modified in the last 7 days. For each file, apply the Content Mining Prompt (see below). Use the Notion MCP to write every extracted block directly into my "Content Blocks" database. After the run, move processed files into /Users/YOURNAME/Documents/Content-Inbox/Processed/. Reply to me with a summary listing the filenames processed and the number of blocks added per file. --- CONTENT MINING PROMPT --- [paste Prompt 4 from above here]
The final step. This task pulls blocks from your Notion database that are ready to produce, matches each block to the right custom skill (from Step 2), and drafts a polished piece. Run weekly on a cadence that matches your posting schedule - most clients do this the day before publishing.
Claude Cowork Recommended
The drafting task runs automatically, pulls from Notion, and writes drafts into a folder or document you specify.
ChatGPT or Other AI
Run the prompt manually, pasting in one Notion block at a time. Takes a minute per draft.
If you're on Claude Cowork (recommended)
- Notion connection already done from Step 4. Skip to the next line.
- Create a drafts folder on your computer (e.g.,
~/Documents/Content-Drafts/). Each draft will land here as a markdown file ready to copy-paste. - Open Claude Desktop → Schedule sidebar → New Task. Name it "Weekly Content Drafting."
- Set frequency: Weekly, Sunday 6pm (the day before your Monday posts go out).
- Paste the Drafting Prompt below as the task instructions, with your drafts folder path and database name filled in.
- On the first run, approve the Notion-read and file-write permissions as "Always allow."
If you're on ChatGPT or another AI (manual run)
- Open your Content Blocks database in Notion
- Pick a block that's Scripted and has a Format Assignment
- Copy every field value into a new AI conversation (with brand docs uploaded)
- Paste the Drafting Prompt below, followed by the block fields
- The AI produces a ready-to-post draft
You are my Content Drafter. Your job is to pull a content block from my Notion Content Blocks database and produce a polished, ready-to-post piece using the matching custom skill for that format. --- INPUT I will either paste a single block below, OR (if running as a Claude Cowork scheduled task) you will read blocks directly from my Content Blocks database where Status = Scripted AND Format Assignment is not Unassigned. Process up to 5 blocks per run unless I tell you otherwise. Each block has at minimum: - Topic / Angle - Core Insight - Verbal Hook (if video) - The Actual Story (real excerpt) - Format Assignment - Best Platform - Script Framework - Content Type --- PROCESS 1. Read the block. Identify the Format Assignment (e.g., FB TOF Post, FB MOF Post, IG Reel, Email, Carousel, YouTube, B-roll Caption). 2. Use the matching custom skill I've installed for that format to draft the piece. The skill handles voice, structure, and rules - you apply the block's content to the skill's template. 3. If I haven't installed a skill for this format yet, draft using my Authentic Voice Guide patterns as a fallback AND flag at the end that I should build this skill next (see Step 2). 4. Use ONLY real details from the block's "Actual Story" field. Never fabricate quotes, numbers, names, client outcomes, or timeline specifics. If the block is thin, tell me what's missing instead of padding. 5. Match the Best Platform's algorithm rules: - Facebook: no external links in post body (put them in the first comment and flag this) - Instagram Reels: caption under 150 characters when possible - Email subject lines: follow my established email subject format (if my Authentic Voice Guide specifies one) - YouTube: hook in the first 5 seconds, not later --- OUTPUT FORMAT For each block drafted, produce: === [FORMAT LABEL, e.g. "FB TOF POST" or "IG REEL CAPTION"] Block ID: [from Notion] Topic: [from block] --- [DRAFT - ready to copy-paste] --- [Character count or length metric] NOTES: - [Any flags: missing info, client initial needed, placeholder in draft, suggested next action] - [If no skill installed for this format, flag: "Build a [format] skill next using Step 2"] - [If the block would benefit from a specific companion piece, suggest it] === --- HARD RULES - No invented quotes, numbers, names, or client outcomes. Ever. - If the block is thin, tell me what to add - don't pad it to feel full. - Match the installed skill exactly - if my FB MOF skill bans "Here's the thing," do not use that phrase. - One draft per block. If the block is strong enough for multiple formats, flag it at the end - don't silently produce multiple versions unless I ask. - If running as a Claude Cowork scheduled task, write each draft to my output folder at [YOUR DRAFTS FOLDER] with a filename format of [YYYY-MM-DD]-[format]-[topic-slug].md and notify me with a summary when done. Ready. Paste a single block below, or proceed to read from Notion.
Cowork task configuration (copy into your scheduled task)
Once Notion is connected and your drafts folder exists, paste this as the task prompt (replace the path with yours):
Use the Notion MCP to read my "Content Blocks" database. Filter for blocks where Status = "Scripted" AND Format Assignment is not "Unassigned" AND Week Assigned is within this week. Process up to 5 blocks per run. For each block, apply the Content Drafting Prompt (see below) using the installed skill that matches the Format Assignment. Save each draft as a markdown file into /Users/YOURNAME/Documents/Content-Drafts/ with filename format YYYY-MM-DD-[format-slug]-[topic-slug].md. After the run, update each processed block's Status in Notion to "In Production." Reply to me with a summary listing each draft filename and any flags (missing info, skill not installed, etc). --- CONTENT DRAFTING PROMPT --- [paste Prompt 5 from above here]