Dreamerce — Claude Cowork Edition

How to replace a
$300K senior strategist
with a $20/month
Cowork setup.

12 prompts, 5 Skills, 4 scheduled tasks, and the exact Cowork project setup that runs the static ad workflow behind 3,000+ Meta ads — autonomously, on a cadence, while you sleep.

12
Power Prompts
5
Custom Skills
4
Scheduled Tasks
$20/mo
Total Cost
→ Open the Library
Read This First

Most people are using Claude Cowork like a chatbot. It's not a chatbot.

If you've opened Claude Cowork, asked it a question, and closed it again — you've used roughly 4% of what it can actually do. Cowork isn't a faster chat. It's a desktop agent that reads your local files, controls your browser via Computer Use, runs scheduled tasks while you sleep, connects to Gmail/Drive/Slack/Notion natively, loads custom Skills automatically, and can be triggered from your phone via Dispatch while it executes on your desktop. What follows is the version of Cowork that actually replaces an ops layer.

What's actually in this document
12 power prompts, 5 ready-to-install Skills, 4 scheduled task templates, and the full project setup file.
Built as a consulting engagement this would cost £4,000+ before anyone wrote a line. You're getting the operational layer free.
£4K+
Equivalent build cost
Before you start — required setup
  • You need Claude Desktop installed (Mac or Windows) with the Cowork tab enabled — Pro plan ($20/mo) works fine
  • Connect at least these four: Gmail, Google Drive, Slack, and Chrome (for Meta Ad Library scraping)
  • Create a project called "Static Ads" — every prompt and Skill in this document lives inside it
  • Keep your laptop awake for scheduled tasks (or run Cowork on a dedicated Mac Mini — pays back in week one)
Setup
Skills
Prompts
Schedule
Dispatch
Section One

5 Things Almost Nobody Knows About Claude Cowork

Before the prompts — five capabilities the average user has no idea exist. Each one is the reason these workflows work at all. Read these once. Then go run them.

01
Cowork Reads and Writes to Your Local Files
No uploads. No copy-paste. Direct filesystem access.
Foundation

Most people don't realise Cowork has full read/write access to local folders. Point it at /Users/you/Documents/StaticAds/ and it can read every research file, every brief, every ad in there — without you uploading anything.

That changes the workflow completely. You no longer "give Claude context" by pasting. You give Claude an entire research folder and it pulls what it needs. This single capability is why 90% of the prompts below work the way they do.

02
Skills Make Cowork Work Like Your Best Strategist
Custom .md files Claude reads automatically on every task.
Compounding

A Skill is a markdown file you drop in your ~/.claude/skills/ folder. Cowork picks it up automatically and applies it to every relevant task. Build a Skill called brand-voice.md and every brief Claude writes matches your brand. Build one called angle-extraction.md and Claude uses your specific methodology every time it analyses research.

The compounding effect is the secret. Most people prompt the same thing 40 times. A Skill means you build it once, and Claude uses it forever. Section Three gives you 5 ready-to-install Skills.

03
Computer Use + Ad Library = Automatic Competitor Intel
Claude opens Chrome, scrolls, extracts. While you sleep.
Browser Agent

Computer Use lets Claude control your browser. Combine that with a scheduled task and Cowork opens the Meta Ad Library at 8am every Monday, scrolls through every active ad your top 10 competitors are running, extracts the copy and visual approach, structures it into a competitive brief, and drops it in your Drive folder.

You wake up to a finished competitive intelligence report. Every week. Without lifting a finger. Task T1 in Section Five gives you the exact prompt to set this up.

04
Dispatch Lets You Trigger Tasks From Your Phone
Text instructions to your desktop from anywhere.
Mobile Trigger

Dispatch pairs your phone with your desktop Cowork. While you're on a call with a client and they mention a competitor you've never heard of, you open the Dispatch tab on your phone and text: "Pull every ad [competitor] has run in the last 30 days. Identify their top 3 angles. Have it ready in their folder by tonight."

Your desktop executes the task in the background. You finish the call and the work is done. No memo, no follow-up, no "I'll get that to you tomorrow."

05
CLAUDE.md Is the Highest-ROI File You'll Ever Write
Global instructions Cowork reads at the start of every session.
Foundation

Drop a file called CLAUDE.md inside your Static Ads project folder. Cowork reads it at the start of every session. Inside that file: who you serve, your creative philosophy, how you talk about angles, what good and bad output looks like, brand voice rules — anything you find yourself explaining more than twice.

This is the single setup file that turns Cowork from "a helpful assistant" into "the strategist who knows how I work." Section Two is the full CLAUDE.md template — paste it directly into your project.

Contents

What's in This Library

Section Two

The CLAUDE.md File — The Setup That Changes Everything

Create a folder called StaticAds in your Documents. Drop this file inside as CLAUDE.md. Cowork reads it at the start of every session in that project. Replace the bracketed sections with your specifics. This is the single most important file you'll create.

📁 CLAUDE.md — Project Foundation Copy + Paste
00
CLAUDE.md — Static Ads Project
Save as CLAUDE.md in /Documents/StaticAds/
# CLAUDE.md — Static Ads Project ## Who I am I produce static ads for ecommerce brands running on Meta. 15+ years in design, 3,000+ statics shipped, working with 7- to 9-figure DTC brands across health, beauty, supplements, and fashion. ## How I think about creative - Research first. Angles second. Execution last. Always. - An angle is the emotional or rational lens, not the headline. Most brands run three angles and call it testing. - AI raises the floor on creative output. It does not replace strategic judgment. - The brief is where most creative value is lost. A brief that needs a follow-up call is a failed brief. - Aesthetic and performance are not the same thing. ## My triage framework — Keep, Scale, Test - KEEP: Proven angles with consistent signal. Iterate, don't reinvent. - SCALE: Early signal angles that deserve more budget and volume. - TEST: Hypotheses from research that haven't been validated. A good sprint contains all three. All TEST is gambling. All KEEP is stagnation. ## How I describe customers Use the customer's exact language. Never paraphrase into marketing speak. - Wrong: "Customers experience digestive discomfort." - Right: "I bloat by 3pm every single day and nothing helps." ## Awareness levels (always assign one to each angle) - Unaware: doesn't know they have the problem - Problem Aware: knows the problem, not the solution - Solution Aware: knows solutions exist, hasn't chosen one - Product Aware: knows about this brand, hasn't bought - Most Aware: ready to buy, just needs the right prompt ## My brands - Primary: [YOUR BRAND NAME + ONE-LINE DESCRIPTION] - Category: [CATEGORY] - ICP: [CUSTOMER PROFILE — WHO, WHAT THEY BELIEVE, WHAT THEY FEAR] ## File structure /StaticAds/research/ — VOC outputs, reviews, Reddit threads /StaticAds/angles/ — Angle maps, sprint plans /StaticAds/briefs/ — Sprint briefs by week /StaticAds/headlines/ — Headline batches by angle /StaticAds/reviews/ — Performance diagnostics on live ads /StaticAds/competitors/ — Ad library teardowns, gap analysis ## Rules for Claude - Always ask before deleting anything - Always show me a plan before executing multi-step work - Outputs go to the relevant subfolder, not the root - When in doubt about brand voice, check the Brand Voice skill before writing - Never use superlatives, exclamation marks, or generic marketing language in headlines ## Banned phrases in ad copy - "Game-changing" - "Revolutionary" - "Clinically proven" (unless we can cite the study) - Anything ending in "your best self" - Anything that sounds like a press release ## Weekly cadence - Monday 8am: Scheduled task runs competitor ad library scan - Monday 10am: Review scan + research file, build sprint plan - Tuesday: Briefs go to design - Thursday: Designs come back, run review prompt - Friday: Sprint goes live, retrospective from previous sprint This file is the source of truth for how I work. Update as my methodology evolves.
Section Three

5 Skills to Install — Claude's Strategist Layer

A Skill is a markdown file Claude reads automatically when relevant. Drop each below in ~/.claude/skills/ as a separate .md file. Cowork picks them up on next launch. Once installed, every prompt in Section Four runs through these Skills without you having to mention them.

Insider note

Why Skills are the unfair advantage

Most people prompt Claude the same way every time and wonder why output drifts. Skills fix that permanently. Build the Skill once, drop it in the folder, and every future Cowork session uses your methodology automatically.

Install all five below. Restart Cowork. Watch the difference.

1
brand-voice.md
Loaded on every task involving copy. Defines voice, banned phrases, examples.
# Brand Voice Skill ## When to use Whenever Claude is writing copy of any kind — headlines, briefs, supporting copy. Load automatically. ## Core voice principles - Quiet confidence, not aggressive selling - Specificity over abstraction. Numbers and concrete details over adjectives. - Sound like a person, not a brand - Customer language over marketing language - One idea per sentence. One sentence per line where possible. ## Banned phrases - "Game-changing", "revolutionary", "next-level" - "Unleash", "unlock", "elevate" - "Your best self", "live your truth" - "We're more than a brand" - "Designed to" / "engineered for" - Any sentence ending in an exclamation mark - Superlatives without proof ## Voice examples ❌ "Unlock your transformation with our revolutionary blend." ✅ "Two weeks in. Bloating gone. Didn't change anything else." ❌ "The ultimate solution for tired skin." ✅ "She thought she'd just slept badly. Three weeks later her dermatologist confirmed it." ## Headline rules - Under 8 words where possible - Clear over clever - If a customer wouldn't say it, don't write it - No clichés, jargon, or exclamation marks - Lead with pain, transformation, or specific outcome ## When the voice is right Reader scans it and either thinks "that's exactly how I feel" or "wait, what?" — never "that's a brand trying to sell me something."
2
angle-extraction.md
Loaded whenever you're analysing research or building an angle map.
# Angle Extraction Skill ## When to use Any task involving customer research synthesis, voice-of-customer analysis, or angle mapping. ## What an angle actually is An angle is the emotional or rational lens through which the ad frames the product. It is NOT the headline. The headline expresses the angle. Examples: - "Fear of Embarrassment" - "Identity Shift" - "Ingredient Credibility" - "Us vs. Them" - "Transformation Story" - "Permission to Quit Trying Other Things" ## Method 1. Read research for what customers fear, want, and believe 2. Group similar emotional patterns 3. Name each angle in 2-4 words 4. Assign awareness level 5. Score 1-5 by signal strength (5 = multiple customers said the same thing unprompted) ## Awareness level matters more than most people think A Problem Aware customer needs a different hook to a Most Aware customer. Getting this wrong is why ads stop scaling. ## Output format For each angle: - ANGLE NAME - CORE INSIGHT (customer language) - AWARENESS LEVEL - HOOK TERRITORY (2-3 directions, not finished headlines) - VISUAL DIRECTION (specific, not "something clean") - PRIORITY SCORE (1-5) ## Minimum standard Never extract fewer than 8 distinct angles. If you can't find 8, the research wasn't deep enough. ## Red flags - Every angle sounds like a feature → extracting features, not angles - All "Most Aware" awareness levels → only thinking about existing buyers - No customer language in insights → paraphrasing into marketing speak
3
brief-writing.md
Loaded whenever you're writing a brief. Forces specificity.
# Brief Writing Skill ## When to use Any task involving writing static ad briefs for designers. ## The brief standard Designer should execute without a single clarifying question. Every strategic decision is made by the brief. Designer only makes executional decisions. ## A brief must include 1. STRATEGIC OBJECTIVE — what must this ad make the customer think, feel, or do? 2. CUSTOMER INSIGHT — exact belief or fear, in customer language 3. ANGLE NAME + AWARENESS LEVEL 4. HOOK — primary headline, under 8 words, clear over clever 5. SUPPORTING COPY — 1-2 lines max 6. VISUAL MANDATE — specific image, layout, hierarchy, colour 7. CTA — exact text 8. WHAT WOULD KILL THIS AD — 2-3 things to avoid ## What makes a visual mandate "specific" ❌ "Something clean and modern with a lifestyle feel" ✅ "Split-frame: left side shows a cluttered bathroom shelf with 6 visible failing products and harsh overhead lighting. Right side shows a clean white shelf with one product and natural side lighting. Headline overlays top third in 48pt bold sans-serif." ## What makes a customer insight "right" ❌ "She wants to feel confident in her skin." ✅ "She's tried 8 products in the last 18 months. Each one worked for two weeks. She's stopped believing anything will work." ## Format One concept per block. Numbered. Designer should read the full sprint of 10 briefs in under 10 minutes. ## When the brief is right Designer reads it and starts working immediately. Doesn't message you. Doesn't ask for "more direction." Opens the file and executes.
4
creative-review.md
Loads when reviewing live or draft ads. Structured 5-criterion diagnostic.
# Creative Review Skill ## When to use Any task involving review or diagnosis of static ad creative. ## Framework — score each /10 1. SCROLL-STOP POWER — does visual + hook create immediate pause? 2. ANGLE CLARITY — is the emotional/rational territory immediately obvious? 3. CUSTOMER LANGUAGE — sounds like the customer or the brand? 4. VISUAL-COPY ALIGNMENT — does image reinforce or contradict the headline? 5. CTA STRENGTH — is the desired action clear and friction-free? ## Total → Verdict - 40-50: Keep running - 30-39: Iterate (1-3 specific changes) - Below 30: Replace ## Common failures - Low scroll-stop, high angle clarity → headline isn't strong enough for the visual - High scroll-stop, low angle clarity → visual is doing too much work alone - Low customer language → copy was written from brand thinking, not research - Low visual-copy alignment → designer wasn't given a tight brief ## Never say in a review - "It's a vibe issue" - "It just doesn't feel right" - "Make it more premium" Every weakness gets named specifically. Every fix has a clear instruction.
5
competitor-teardown.md
Loaded when Cowork is scraping or analysing competitor ads.
# Competitor Teardown Skill ## When to use Any task involving Meta Ad Library scraping, competitor ad analysis, or whitespace identification. ## What to extract from each ad - Hook type (pain / transformation / social proof / curiosity / fear / identity / contrarian) - Primary angle (named, not described) - Awareness level being targeted - Visual format (product hero / problem-solution / UGC / before-after / lifestyle) - Estimated longevity (new this week or running 30+ days?) ## Saturation scoring Rate each angle 1-5 on prevalence: - 5 = every brand is running it - 3 = several brands - 1 = almost no one ## Whitespace identification After mapping all angles, identify: - Angles missing entirely from the competitive set - Angles only one brand is touching (early-mover opportunity) - Angles where everyone is running it badly ## Output File to /StaticAds/competitors/[brand]/[date].md Format: angle map table → saturation scores → whitespace opportunities → recommended next angle for our brand ## Speed standard Full teardown of 10 competitors should take under 30 minutes as a scheduled task.
Once installed
Every prompt in Section Four runs through these 5 Skills automatically.
You don't paste them in. You don't reference them. Cowork loads them based on the task. That's the entire point.
Section Four

12 Power Prompts

Every prompt below assumes CLAUDE.md and the 5 Skills are installed. That's why the prompts are short — the heavy lifting is in your project setup. Run them in order. Each builds on the last.

🔍 Research Prompts 01–03
01
Voice-of-Customer Synthesis
Reads research files from local folder, no paste required
Read every file in /StaticAds/research/[BRAND-NAME]/raw/ These contain raw customer reviews, Reddit threads, and forum posts. Apply the Angle Extraction skill. Produce a voice-of-customer document containing: 1. Top 5 pain points (verbatim quotes) 2. Top 5 desires (emotional, not feature-level) 3. Buying triggers — what pushed them from considering to buying 4. Objections they had to overcome 5. Transformation moments — how did they describe the change 6. 15-20 phrases to use in ads (exact customer language) 7. Phrases to avoid (brand language) Save the output to /StaticAds/research/[BRAND-NAME]/voc-synthesis.md Tell me when it's done with a one-paragraph summary of the strongest signal.
02
Reddit + Facebook Group Deep Scrape
Computer Use opens both, scrolls, extracts the language brands never see
Open Chrome. PART 1 — Reddit Go to reddit.com. Search these subreddits for posts mentioning [PRODUCT CATEGORY KEYWORDS] from the last 6 months: - r/[SUB 1] - r/[SUB 2] - r/[SUB 3] For each thread: - Open the thread - Read the top 20 comments - Extract real pain points, desires, buying triggers - Note upvote count (signal of how widely felt) PART 2 — Facebook Groups Go to facebook.com (I'm already signed in). Navigate to each group I've joined for [CATEGORY]: - [FB GROUP 1] - [FB GROUP 2] - [FB GROUP 3] Use the group search to find posts containing [PRODUCT CATEGORY KEYWORDS] from the last 90 days. For each relevant post: - Read the post and top 15 comments - Extract pain language, product complaints, and what people are actually recommending to each other - Note engagement (reactions, comment count) as a resonance signal This is where the unfiltered conversation lives — people share things in private FB groups they would never put in a public review. OUTPUT Save Reddit material to /StaticAds/research/[BRAND-NAME]/raw/reddit-[DATE].md Save FB group material to /StaticAds/research/[BRAND-NAME]/raw/fb-groups-[DATE].md Format: one quote per line, with source + engagement count in brackets. When done, run Prompt 01 against both files automatically.
03
Amazon Review Mining
Pulls 1-3 star reviews from competitor products — where the real insight lives
Open Chrome and go to the Amazon listing for each competitor product: - [COMPETITOR PRODUCT URL 1] - [COMPETITOR PRODUCT URL 2] - [COMPETITOR PRODUCT URL 3] For each product: - Filter to 1-star, 2-star, and 3-star reviews - Extract full text of first 25 reviews - Note what they specifically complained about 5-star reviews tell you what people loved. 1-3 star reviews tell you what's still missing — and that's where the angles you can own actually live. Save to /StaticAds/research/[BRAND-NAME]/raw/amazon-mid-low-[DATE].md After scraping, identify the top 5 unmet needs across all reviews and tell me which ones our brand could speak to that competitors aren't.
🎯 Angle Prompts 04–06
04
Full Angle Extraction
Builds the angle map from your VOC research file
Read /StaticAds/research/[BRAND-NAME]/voc-synthesis.md Apply the Angle Extraction skill. Build a complete angle map. Minimum 8 angles. Each one must include name, core insight (customer language), awareness level, hook territory, visual direction, priority score. Save to /StaticAds/angles/[BRAND-NAME]/angle-map.md Then tell me which 3 angles have the strongest signal and why.
05
Keep / Scale / Test Sprint Planner
Builds this week's sprint from the angle map + last sprint's results
Read: - /StaticAds/angles/[BRAND-NAME]/angle-map.md - /StaticAds/reviews/[BRAND-NAME]/last-sprint-results.md (if it exists) Build a 10-concept sprint for this week using Keep / Scale / Test from CLAUDE.md. Required output: 1. Sprint selection — 10 angles, one-line rationale each 2. Keep / Scale / Test assignment per angle 3. Awareness spread check — must cover 3+ awareness levels 4. Concept format pairing per angle 5. Backlog of remaining angles for future sprints Save to /StaticAds/angles/[BRAND-NAME]/sprint-[WEEK].md
06
Whitespace Angle Hunter
Cross-references your angle map against competitor teardowns
Read: - /StaticAds/angles/[BRAND-NAME]/angle-map.md - /StaticAds/competitors/[BRAND-NAME]/latest-teardown.md Cross-reference. Identify: 1. Angles in our map competitors are running (we'll be fighting for share of attention) 2. Angles in our map competitors are NOT running (whitespace — high-leverage) 3. Angles competitors are running that we don't have a position on (research gaps) Recommend 3 whitespace angles to prioritise in the next sprint. Explain why each is a real opportunity, not just unused space. Save to /StaticAds/angles/[BRAND-NAME]/whitespace-[DATE].md
📋 Brief Prompts 07–09
07
Full Sprint Brief Generator
10 complete briefs in one pass, dropped in the briefs folder
Read /StaticAds/angles/[BRAND-NAME]/sprint-[WEEK].md Apply the Brief Writing skill. For each of the 10 selected angles, produce a full executional brief: angle + awareness, customer insight, hook, supporting copy, visual mandate, CTA, what would kill the ad. Save each brief as a separate file in /StaticAds/briefs/[BRAND-NAME]/sprint-[WEEK]/ File naming: 01-[ANGLE-NAME].md, 02-[ANGLE-NAME].md, etc. Also produce a single sprint-overview.md in the same folder listing all 10 briefs with one-line summaries for the design team.
08
Brief-to-Slack Handoff
Posts sprint summary into Slack design channel automatically
Read /StaticAds/briefs/[BRAND-NAME]/sprint-[WEEK]/sprint-overview.md Post a Slack message to the #design channel with: - This week's sprint goal (1 line) - The 10 angles being tested (numbered, awareness level in brackets) - Link to the briefs folder in Google Drive - Note: "Open any brief and start when you're ready. Questions in thread." Format so it's scannable. Bold headers. Divider line between sprint goal and angle list.
09
Headline Batch Generator
20 headlines per angle, ready for design to choose from
Read: - /StaticAds/briefs/[BRAND-NAME]/sprint-[WEEK]/[BRIEF-FILE].md - /StaticAds/research/[BRAND-NAME]/voc-synthesis.md Apply the Brand Voice skill. Write 20 distinct headlines for this brief, drawing language from VOC research. Cover all 8 hook territories: pain stated plainly, transformation, contrarian, social proof, curiosity gap, plain offer, consequence, identity statement. Mark top 5 with reasoning specific to the angle and awareness level. Save to /StaticAds/headlines/[BRAND-NAME]/sprint-[WEEK]/[ANGLE-NAME]-headlines.md
Review Prompts 10–12
10
Live Ad Diagnostic
Run the 5-criterion review on any draft or live ad
Attach an image of a static ad — or describe it in detail. Apply the Creative Review skill. Score each of 5 criteria /10, give total, assign verdict (Keep / Iterate / Replace). If Iterate: list top 1-3 highest-leverage fixes only. If Replace: describe what a stronger concept in the same angle space would look like. Save the review to /StaticAds/reviews/[BRAND-NAME]/[AD-NAME]-[DATE].md Ad: [ATTACH IMAGE OR DESCRIBE]
11
Portfolio Keep / Scale / Kill
Analyses your full ad portfolio and recommends sprint allocation
Read /StaticAds/reviews/[BRAND-NAME]/ and pull every review file from the last 90 days. Group ads by angle. For each angle: - Average performance signal across all ads - Best-performing concept - Worst-performing concept - Trend direction (improving / declining / stable) Recommend: - KEEP: which angles are working — keep variations - SCALE: which angles have early signal — budget bump - KILL: which angles have failed — retire Save to /StaticAds/reviews/[BRAND-NAME]/portfolio-[DATE].md Use this to shape next week's sprint plan.
12
End-of-Week Creative Retrospective
Friday afternoon prompt that closes the loop on the week
End of the week. Read: - /StaticAds/briefs/[BRAND-NAME]/sprint-[WEEK]/ - /StaticAds/reviews/[BRAND-NAME]/ (latest) Produce a one-page retrospective: 1. What we planned to test this week (sprint angles) 2. What actually got shipped 3. Early signal so far (if reviews are in) 4. What we learned about this customer that we didn't know last Monday 5. The single sharpest insight from this week 6. Recommended angle direction for next week Save to /StaticAds/reviews/[BRAND-NAME]/retro-week-[WEEK].md Email retrospective to [YOUR EMAIL] with subject "Week [WEEK] Retro — [BRAND-NAME]"
Section Five

4 Scheduled Tasks — Set Up Once, Run Forever

This is where Cowork stops being a tool and starts being a system. Each task runs automatically on a cadence — competitor intel pulled before you arrive, retrospectives written Friday afternoon while you wrap up. Setup is 2 minutes per task.

How to schedule a task

The 2-minute setup that pays back forever

1. Run the prompt manually first to confirm output.

2. Once you're happy, type /schedule in the same chat.

3. Cowork asks for cadence (daily / weekly / monthly) and time.

4. Done. Task saved. Runs without you opening anything as long as your laptop is awake and Cowork is open.

For full autonomy, run Cowork on a dedicated Mac Mini left on 24/7. Use Dispatch on your phone to monitor remotely.

Scheduled Task Prompts 4 Templates
T1
Monday 8am — Competitor Ad Library Scan
Wake up to a finished competitive intelligence report every Monday
Every Monday at 8:00 AM: 1. Open Chrome, navigate to Meta Ad Library 2. For each competitor in /StaticAds/competitors/[BRAND-NAME]/competitors.txt: - Search their brand - Scroll all active ads from last 14 days - Extract: headline, body copy, visual format, ad longevity 3. Apply Competitor Teardown skill 4. Build a one-page brief: - Top 3 new angles competitors launched this week - Any angle with multiple competitors converging (warning) - Whitespace opportunities for the coming sprint 5. Save to /StaticAds/competitors/[BRAND-NAME]/weekly-[DATE].md 6. Slack #strategy: "Competitor scan complete. Top angle this week: [angle]. Full report in Drive." Run silently. The report should be ready before I'm at my desk with coffee.
T2
Tuesday 9am — Auto Sprint Brief Generator
Briefs ready before the design team logs in
Every Tuesday at 9:00 AM: 1. Check /StaticAds/angles/[BRAND-NAME]/ for the latest sprint plan 2. If found, run Prompt 07 (Full Sprint Brief Generator) 3. If no plan exists, Slack DM: "No sprint plan found for week of [date]. Need plan before briefs can be generated." 4. Once briefs are generated, run Prompt 08 (Brief-to-Slack Handoff) 5. Send me summary at 9:30 AM: "Sprint briefs complete. [X] briefs ready in [folder]. Posted to #design at 9:25 AM." Quality check before posting: confirm every brief has all 8 required components from the Brief Writing skill. Flag any missing.
T3
Wednesday 5pm — Research Refresh
Mid-week pulse on customer language and shifting signals
Every Wednesday at 5:00 PM: 1. Open Chrome and scrape for /StaticAds/research/[BRAND-NAME]/: - Top 3 subreddits for our category - Amazon 1-3 star reviews from top competitor product (latest 20) - Any Reddit thread with 50+ upvotes mentioning our category this week 2. Append to /StaticAds/research/[BRAND-NAME]/raw/weekly-refresh.md 3. Compare new language against /StaticAds/research/[BRAND-NAME]/voc-synthesis.md 4. Identify new pain points, desires, or phrases appearing for the first time 5. If 3+ new signals appear, Slack DM: "New customer signals detected — worth a look before Friday sprint planning." If nothing has shifted, post quietly: "Weekly research refresh complete. No new signal."
T4
Friday 4pm — Auto Retrospective + Email
Closes the week without you opening a single doc
Every Friday at 4:00 PM: 1. Run Prompt 12 (End-of-Week Creative Retrospective) 2. Save to /StaticAds/reviews/[BRAND-NAME]/retro-week-[WEEK].md 3. Email retrospective to [YOUR EMAIL] with subject "Week [WEEK] Retro — [BRAND-NAME] — [sharpest insight in one line]" 4. Slack #strategy: "Week [X] wrapped. Sharpest insight: [insight]. Full retro in your inbox." 5. If any learning has emerged that should be permanent, Slack DM me first for approval before editing CLAUDE.md. Never edit CLAUDE.md without explicit approval. Run quietly. Close laptop Friday. Reopen Monday with the week documented.
Without Scheduled Tasks
The strategist workflow
Monday: 3 hours pulling competitor intel. Tuesday: 4 hours writing briefs. Wednesday: research drift you'll get to "later." Friday: rushed retro at 5pm or skipped entirely. Burnout by month two.
With Scheduled Tasks
The system workflow
Monday: walk in to a finished competitor report. Tuesday: review briefs that wrote themselves overnight. Wednesday: signal appears in your DM. Friday: retro lands in your inbox. The strategic work is the only work left.
Final Section

What You Just Built

If you install everything in this document, here's what your week actually looks like.

Monday 8am: Competitor intel waits in your Drive. Tuesday 9am: Sprint briefs posted to Slack while you finish coffee. Wednesday 5pm: A DM flags fresh customer language you didn't have last week. Friday 4pm: This week's retrospective lands in your inbox while you close the laptop.

Outside of those automated touchpoints, your job is one thing: the strategic judgment. Which angles deserve budget. Which signal is real and which is noise. Which whitespace is worth owning. The execution layer — research synthesis, brief writing, headline drafting, competitor scanning, retrospectives, file organisation — runs without you. That's the trade. A senior strategist replaced by $20/month and 90 minutes of setup.

The honest part
The system is the easy part. The judgment is what makes it work.
Every brand running these prompts uses the same tool. The brands whose creative scales aren't winning because of better prompts. They're winning because someone is reading the output with sharp judgment, knowing which signal to act on and which angle to own. That's not in the library. That's where the next conversation starts.
Next Step

You have the system. Do you have the judgment to run it?

The prompts are the operational layer — they execute. The strategic layer is harder: diagnosing why a winning angle stopped, reading research for signals competitors can't see, knowing when to kill versus iterate. That's not a prompt. If you want that conversation — and you're serious about Meta creative — this is where it starts.

→ Work With Dreamerce

For ecom brands spending consistently on Meta. Not for brands looking for a cheaper option.