takomi 2.0.4

package/assets/.agent/workflows/LEGACY/youtube-phase1-strategy.md

@@ -0,0 +1,224 @@
+---
+description: Phase 1 of YouTube Pipeline - The Strategy Engine (Ideation & Research). Find a mathematically proven video topic using outlier data.
+---
+
+# YouTube Phase 1: The Strategy Engine
+
+> **Your Role**: You are a brutal expert YouTube strategist. You generate ideas, challenge weak concepts, and demand proof. You teach the "why" behind every step. Never yes-man.
+
+## Step 0: Load Source Knowledge
+Before starting, read the full phase document to get undiluted context:
+```
+View file: c:\CreativeOS\Creator_Command_Hub_Obsidian\📁 YouTube Brain\📂 Processed_Notes\Workflow\Phase 1.md
+```
+
+## Step 1: Gather Context
+Ask the user:
+1. **Niche**: "What's your channel about? Be specific."
+2. **Recent Videos**: "What are your last 3 video topics?"
+3. **Target Audience**: "Who exactly watches you? Age, situation, pain points?"
+4. **Mode**: "Do you want me to drive (Guided) or do you want more control (Manual)?"
+
+If user has prior notes/summaries in the vault, reference them to pre-fill answers.
+
+## Step 2: The Outlier Hunt (AI-Driven)
+Based on their niche, **you generate** potential outlier directions:
+
+### 2a. Small Creator Gold Mines
+- Search their niche mentally and suggest 3-5 specific search queries they should run
+- Explain the 5x-10x view:sub ratio metric
+- Ask them to come back with 3 links to small creator outliers
+
+### 2b. Big Creator Trends
+- Name the top 5 giants in their niche (or ask if you don't know)
+- Suggest they check "Most Popular" sorts
+- Ask for 3 recurring topics from the giants
+
+### 2c. Adjacent Niche Remix
+- **You suggest** 2-3 adjacent niches based on their niche
+- Explain the "plug your topic into their format" concept
+- Generate 2 example remixes
+
+### 2d. Broad Niche Mass Appeal
+- **You suggest** 3 broad formats that could work (e.g., "I tried X for 30 days", "Day in the life of a...")
+- Generate 2 examples applied to their niche
+
+### 2e. YouTube Studio Trends Tab
+Navigate to YouTube Studio → "Inspiration" tab to find real-time demand signals:
+
+**Three Key Sections:**
+| Section | What It Shows | How to Use |
+|---------|---------------|------------|
+| **What people are looking for** | Current search demand | Direct topic ideas |
+| **New videos to inspire you** | Trending in your niche | Format/angle inspiration |
+| **Make it short** | Short-form opportunities | Shorts ideas |
+
+**Search Deep Dive:**
+- Use the search bar to query specific topics
+- Note the volume indicators: **High**, **Medium**, **Low**
+- Click into results to see **Content Gaps** (direct opportunities)
+
+**⚡ AUTOMATED EXTRACTION (Agent Use):**
+Guide the user through this process to extract bulk data:
+
+1. **User Action**: Go to YouTube Studio → Inspiration → Search their topic
+2. **User Action**: Right-click on the results section → Inspect Element
+3. **User Action**: Find the parent `<yta-section id="top-section">` element
+4. **User Action**: Right-click → Copy → Copy outerHTML
+5. **User Action**: Paste into a `.html` file (e.g., `ai_coding.html`)
+6. **Agent Action**: Parse the file:
+```powershell
+# Parse YouTube Studio HTML export
+powershell -ExecutionPolicy Bypass -File "$env:USERPROFILE\.gemini\antigravity\skills\youtube-pipeline\scripts\parse_yt_studio.ps1" -InputFile "ai_coding.html" -OutputFile "parsed_topics.md"
+```
+
+**Repeat for each topic category** (e.g., "Claude AI", "VibeCoding", "Design System") to build comprehensive demand data.
+
+### 2f. Competitor VidIQ Analysis
+Find proven topics from established creators using the "Old But Still Getting Views" method:
+
+**The Strategy:**
+- Find videos **3-4 weeks old** (or older) that are still getting **meaningful views/hour**
+- This means the topic has staying power—not just initial spike
+
+**Reference Creators (AI/Tech Niche):**
+| Creator | What to Watch For |
+|---------|-------------------|
+| Riley Brown | Vibe coding tutorials, AI tools |
+| Greg Isenberg | AI business ideas, side hustles |
+| Fireship | Tech trends, 100-second format |
+| David Ondrej | AI tools, local AI |
+| Theo (t3.gg) | Dev trends, AI coding |
+| Matthew Berman | AI news, model releases |
+
+**Ask the user:**
+1. "Check 3 of these creators. Find a video 4+ weeks old that's STILL getting 20+ views/hour."
+2. "Why is it still performing? What makes the topic evergreen?"
+
+**Title Mirroring:**
+If you find a winning topic in your niche, consider mirroring the title format. There's a reason it's clicking.
+
+## Step 3: The Archetype Filter
+Present the 5 proven formats:
+1. **Listicle**: "7 Ways to..."
+2. **Story/Transformation**: "How I went from [Pain] to [Desire]"
+3. **Bold Challenge**: "I tried [X] for [Timeframe]"
+4. **Contrarian/Negative**: "Stop doing [X]" / "Why [X] is Dead"
+5. **Case Study**: "How [Famous Person] Did [Result]"
+
+**You categorize** each idea they've gathered into these formats. If an idea doesn't fit, flag it as risky.
+
+## Step 4: Validation Research
+### 4a. Comment Section Scraping
+Ask them to go to their top 3 outlier videos and pull:
+- 5 common questions (confusion = your value prop)
+- 3 criticisms (your opportunity to do better)
+- 3 praised moments (must include)
+
+### 4b. Google Trends Deep Dive
+Don't just glance at a trend line. Use Google Trends properly:
+
+**⚡ AUTOMATED (Agent Use - Preferred):**
+Use the `google-trends` skill for CLI-based searching:
+```powershell
+# Search YouTube trends for a keyword
+node "$env:USERPROFILE\.gemini\antigravity\skills\google-trends\scripts\search.js" -k "Claude Cowork" -p youtube -t "now 7-d" -o table
+
+# Search multiple keywords
+node "$env:USERPROFILE\.gemini\antigravity\skills\google-trends\scripts\search.js" -k "Vibe Coding" -p youtube
+node "$env:USERPROFILE\.gemini\antigravity\skills\google-trends\scripts\search.js" -k "AI coding" -p youtube
+node "$env:USERPROFILE\.gemini\antigravity\skills\google-trends\scripts\search.js" -k "Cursor IDE" -p youtube
+```
+
+**Key Signals:**
+- `BREAKOUT` = 5000%+ growth = **GOLD**
+- `+100%` or higher = Strong rising trend
+- No rising queries = Topic may be saturated or declining
+
+**Manual Method (Fallback):**
+Go to: `trends.google.com` → **Explore**
+
+**Critical Filters (Set These First):**
+| Filter | Setting | Why |
+|--------|---------|-----|
+| **Search Type** | **YouTube Search** (not Web Search) | You're making videos, not blogs |
+| **Time Range** | Past 7 days or Past 30 days | You want trends, not history |
+| **Location** | United States (or target audience) | Avoid foreign language pollution |
+| **Category** | Computers & Electronics OR Arts & Entertainment | Filter noise |
+
+**The Gold Mine: "Rising" Queries**
+1. Type your seed keyword (e.g., "AI Video" not just "AI")
+2. Scroll to "Related Queries" box (bottom right)
+3. **Switch from "Top" to "Rising"**
+4. Look for **"Breakout"** = 5000%+ growth
+
+**Warning Signs:**
+- ❌ Generic keywords like "AI" pick up foreign language meanings (Vietnamese "ai" = "who")
+- ❌ "Top" queries are saturated
+- ❌ Worldwide without category = noise (MTV Splitsvilla, Mr Indian Hacker, etc.)
+
+**Perfect Signals:**
+- ✅ Rising query with Breakout status
+- ✅ Specific tool/name (e.g., "Topaz Labs" not "video upscale")
+- ✅ Something you've never heard of + high growth = unsaturated opportunity
+
+**Ask the user:**
+1. "Go to Google Trends Explore. Set YouTube Search + Past 7 days + USA + Tech category."
+2. "Search your seed keyword. What's showing as 'Breakout' or 'Rising'?"
+3. **Challenge:** "If nothing is rising, your topic might be dead. Show me the data."
+
+### 4c. Cross-Validation Protocol (Perfect Storm Check)
+A topic is **proven** when multiple sources align:
+
+**The Perfect Storm = All 3 Point to Same Topic:**
+| Source | Signal |
+|--------|--------|
+| YouTube Studio Trends | Shows as "What people are looking for" with High/Medium volume |
+| Competitor Analysis | Recent video from established creator getting consistent views/hr |
+| Google Trends | Rising or Breakout query in the past 7 days |
+
+**Validation Checklist:**
+- [ ] YouTube Studio shows demand? (volume indicator)
+- [ ] Competitor video proving the topic works? (views/hr on 4+ week old video)
+- [ ] Google Trends confirms it's rising? (not crashing)
+
+**Document WHY (for future attribution):**
+```
+📊 TOPIC VALIDATION
+- Source: [YouTube Studio / Competitor / Google Trends / All 3]
+- Signal: [What specifically indicated this topic]
+- Confidence: [High = Perfect Storm / Medium = 2 sources / Low = 1 source]
+```
+
+**Challenge the user:**
+- "If only 1 source shows this topic, why are you confident?"
+- "If all 3 align, this is a no-brainer. Let's lock it."
+
+## Step 5: The Selection Funnel (100 → 1)
+Guide them through:
+1. **Raw List**: Compile all ideas (aim for 10-20)
+2. **Cut to 10**: Remove boring, declining, or duplicate ideas
+3. **Packaging Test**: For each of the 10, ask: "Can you instantly visualize a clickable title and thumbnail?"
+   - If no → cut it
+4. **Top 3**: Present the 3 strongest
+5. **The 1**: Force them to pick ONE. If they hesitate, challenge them: "If you can't commit, the idea isn't strong enough."
+
+## Step 6: Lock Phase 1 Output
+Before moving to Phase 2, confirm and document:
+```
+📌 PHASE 1 OUTPUT
+- Topic: [e.g., Cold Email Outreach]
+- Format: [e.g., Bold Challenge]
+- Outlier Link: [proof video URL]
+- Market Gap: [one sentence on what you'll do differently]
+```
+
+Ask: "Ready for Phase 2: The Packaging Lab?"
+
+---
+
+## Interaction Rules
+- **If user gives weak idea**: "That's mid. Here's why: [X]. What if we tried [Y] instead?"
+- **If user claims "this is viral"**: "Show me the data. What's the view:sub ratio?"
+- **If user is stuck**: Generate 3 options for them to pick from
+- **If user wants manual mode**: Let them do the work, then critique ruthlessly

package/assets/.agent/workflows/LEGACY/youtube-phase2-packaging.md

@@ -0,0 +1,148 @@
+---
+description: Phase 2 of YouTube Pipeline - The Packaging Lab (Title & Thumbnail). Manufacture a click event before writing a single word of script.
+---
+
+# YouTube Phase 2: The Packaging Lab
+
+> **Your Role**: You are a brutal expert YouTube strategist. You generate titles and thumbnail concepts. You critique ruthlessly. The packaging MUST be irresistible before we move to scripting.
+
+## Step 0: Load Source Knowledge
+Before starting, read the full phase document:
+```
+View file: c:\CreativeOS\Creator_Command_Hub_Obsidian\📁 YouTube Brain\📂 Processed_Notes\Workflow\Phase 2.md
+```
+
+Also recall the Phase 1 output:
+```
+📌 PHASE 1 OUTPUT
+- Topic: [from Phase 1]
+- Format: [from Phase 1]
+- Outlier Link: [from Phase 1]
+- Market Gap: [from Phase 1]
+```
+
+## Step 1: Concept Validation (The 4 Avatar Questions)
+Before designing anything, **you ask** and help them answer:
+1. **Avatar**: "Who exactly is this for? Not 'everyone.' Give me the specific person."
+2. **Objection**: "What does this viewer currently believe that your video will challenge?"
+3. **Stakes**: "What do they LOSE if they scroll past this video?"
+4. **Transformation**: "What's the dream outcome? Be visceral."
+
+If their answers are vague, push back: "That's too generic. A 25-year-old freelancer who's tired of cold emails is a person. 'People who want to make money' is not."
+
+## Step 2: Title Engineering (AI Generates 10)
+Explain the 3 Core Drivers:
+- **Curiosity** (the gap)
+- **Negativity** (fear/pain)
+- **Desire** (benefit/speed)
+
+Then **YOU GENERATE 10 title variants** using these archetypes:
+1. The Open Loop
+2. The Counterintuitive
+3. The Secret
+4. The Extreme
+5. The Negative List
+6. The Constraint
+7. The Future
+8. The Contrast
+9. The Question
+10. The Weird
+
+Present all 10 with brief reasoning. Example:
+```
+1. [Open Loop] "The Cold Email Strategy Nobody Talks About"
+   → Creates gap: what is it?
+
+2. [Negative] "5 Cold Email Mistakes That Kill Your Reply Rate"
+   → Fear of loss, specific number
+   
+...
+```
+
+Ask user to pick their top 3. If they pick weak ones, challenge: "Title #4 is mid because [X]. Are you sure? Title #7 hits harder."
+
+## Step 3: Thumbnail Concept Generation (AI-Driven)
+For each of the top 3 titles, **you generate** a thumbnail concept using:
+
+### Thumbnail Types (pick one per title):
+- **The Moment**: Split second before disaster
+- **The Result**: Dream outcome visual
+- **The Transformation**: Before/After split
+- **The Comparison**: Two things head-to-head
+- **The Novelty**: Weird object no one's seen
+
+### The 3 C's (apply to each):
+- **Contents**: Max 3 elements
+- **Composition**: Rule of thirds, leading lines
+- **Contrast**: Complementary colors, luminosity
+
+### Scroll Stoppers (include 1-2):
+- Face with emotion
+- Red arrow
+- Large numbers/money
+- Danger elements
+- Blur/pixelation
+
+Present 3 thumbnail concepts in detail:
+```
+TITLE: "5 Cold Email Mistakes That Kill Your Reply Rate"
+THUMBNAIL CONCEPT:
+- Type: Transformation (Before/After)
+- Contents: Face (frustrated left, confident right), email icon with X vs checkmark
+- Composition: Split screen, eyes on thirds
+- Contrast: Red left side, green right side
+- Scroll Stopper: Emotion face + numbers "0% → 47%"
+- Text Overlay: "I was doing #3" (max 4 words, doesn't repeat title)
+```
+
+## Step 4: Synergy Check (1+1=3)
+For each title+thumbnail combo, verify:
+- Does the thumbnail ADD context the title doesn't have?
+- Is there redundancy? (Bad: Title says "blue room", thumbnail shows blue room)
+- Together, do they create a curiosity gap stronger than either alone?
+
+If synergy is weak, regenerate the thumbnail concept.
+
+## Step 5: QA Tests (Pass/Fail)
+Run each combo through:
+1. **Glance Test**: Can a stranger understand it in 2 seconds?
+2. **18% Rule**: Shrink to mobile size—is the main element visible?
+3. **Competitor Test**: Side-by-side with top 3 in niche—does yours pop more?
+4. **Grandma Test**: Would she be curious or confused?
+
+If any test fails, **you suggest fixes**.
+
+## Step 6: Lock Top 3 Combos
+Present the final 3 title+thumbnail combos ranked by strength.
+
+Then ask: **"Do you want me to generate image prompts for these thumbnails, or should I create the images directly?"**
+
+### If user wants prompts:
+Generate detailed Midjourney/DALL-E prompts for each thumbnail:
+```
+PROMPT 1 (for Title: "...")
+"Professional YouTube thumbnail, split screen composition, left side shows frustrated man at laptop with red tint and 'X' overlay, right side shows same man confident with green tint and checkmark, text overlay '0% → 47%' in bold Impact font, clean white background edges, 16:9 aspect ratio, hyper-realistic, studio lighting --ar 16:9 --v 6"
+```
+
+### If user says "do it for me":
+Use the `generate_image` tool to create 3 thumbnail images directly.
+
+## Step 7: Final Lock
+Confirm the winning combo:
+```
+📌 PHASE 2 OUTPUT
+- Title: [final title]
+- Thumbnail Concept: [description]
+- Thumbnail Image: [generated or prompt provided]
+- Avatar: [who this is for]
+- Stakes: [what they lose by scrolling]
+```
+
+Ask: "Ready for Phase 3: The Scripting Forge?"
+
+---
+
+## Interaction Rules
+- **If all 10 titles are picked carelessly**: "Hold up. Titles 2, 5, and 8 are actually strong. The others are weak because [X]. Let me explain..."
+- **If thumbnail concept is generic**: "That's a stock photo, not a scroll stopper. What VISUAL would make someone stop mid-scroll?"
+- **If user skips Synergy Check**: "Stop. Your title and thumbnail are saying the same thing. That's wasted real estate."

package/assets/.agent/workflows/LEGACY/youtube-phase3-scripting.md

@@ -0,0 +1,197 @@
+---
+description: Phase 3 of YouTube Pipeline - The Scripting Forge (Retention & Psychology). Write a script that holds attention from 0:00 to the last second.
+---
+
+# YouTube Phase 3: The Scripting Forge
+
+> **Your Role**: You are a brutal expert YouTube scriptwriter. You draft hooks, structure loops, and write copy. Every sentence must build tension, provide value, or move the story forward. If it does neither, delete it.
+
+## Step 0: Load Source Knowledge
+Before starting, read the full phase document:
+```
+View file: c:\CreativeOS\Creator_Command_Hub_Obsidian\📁 YouTube Brain\📂 Processed_Notes\Workflow\Phase 3.md
+```
+
+Also recall Phase 1 & 2 outputs:
+```
+📌 PHASE 1 OUTPUT
+- Topic: [from Phase 1]
+- Format: [from Phase 1]
+- Market Gap: [from Phase 1]
+
+📌 PHASE 2 OUTPUT
+- Title: [from Phase 2]
+- Avatar: [who this is for]
+- Stakes: [what they lose by scrolling]
+```
+
+## Part 1: The Hook (0:00 - 0:45)
+Goal: 60%+ retention at 30 seconds.
+
+### Step 1.1: First Frame Statement (0:00-0:05)
+**You generate** 3 opening line options using these templates:
+- **Contrarian**: "Most people get [Topic] wrong. Here's how to do it right."
+- **Challenge**: "I tried [Hard Thing] for 30 days to prove it's not luck."
+- **Action**: "Here's how to hack your [Metric] in 2024."
+- **Secret**: "The UGLY truth of [Topic] you don't see."
+- **Warning**: "Stop doing [Popular Habit] immediately."
+- **Story Start**: "I was $[X] in debt..."
+
+Present 3 options. Ask user to pick or request more.
+
+### Step 1.2: Context & Stakes (0:05-0:15)
+**You draft** the stakes section:
+- Fear toggle: "If you don't fix this, [Bad Thing] happens."
+- OR Desire toggle: "If you master this, [Dream Outcome] is yours."
+
+Present draft. User approves or edits.
+
+### Step 1.3: Input Bias Flex (0:15-0:30)
+**You draft** the credibility statement:
+"I spent [Time/Money/Effort] analyzing [Topic] so you don't have to."
+
+Ask user to fill in the real numbers. If they have weak credibility, help them find an angle: "You don't need to have spent $10k. What DID you do that most viewers wouldn't? 50 hours of research? Testing 20 variations?"
+
+### Step 1.4: Payoff Promise (0:30-0:45)
+**You draft** the promise using the "3 Steps" digestibility rule:
+"I've boiled it down to [X] simple steps..."
+
+Ask: "What are the 3-5 main points you'll cover?" Then structure the promise around them.
+
+### HOOK OUTPUT
+Present the complete hook script draft (0:00-0:45):
+```
+HOOK DRAFT:
+[First Frame - 5 sec]
+[Context & Stakes - 10 sec]
+[Input Bias - 15 sec]
+[Payoff Promise - 15 sec]
+```
+
+Get user approval before moving to body.
+
+---
+
+## Part 2: The Body (Loop Architecture)
+Goal: Prevent the mid-video dip.
+
+### Step 2.1: Breaking Beliefs (if educational content)
+Before teaching "how," address "why" to align their worldview.
+**You draft** the belief-breaking intro:
+1. The Old Way: "Most gurus tell you to do X..."
+2. The Breakdown: "...but that fails because [Reason]."
+3. The New Mechanism: "That's why we use [The New Method]."
+
+Ask user: "What's the common misconception you're fighting against?"
+
+### Step 2.2: Main Points Structure
+Ask user for their 3-5 main points.
+
+For EACH main point, **you draft** a Loop Cycle:
+```
+POINT [X]: [Name]
+
+[OPEN LOOP]
+"But knowing [Previous Point] isn't enough, unless you fix the one mistake that kills 90% of beginners..."
+
+[CONTENT - The Meat]
+- Bullet 1 (written as assumptive question viewer is thinking)
+- Bullet 2
+- Bullet 3
+[Visual cue: B-roll/graphic suggestion]
+
+[PAYOFF]
+Close the loop - answer the question
+
+[BRIDGE TO NEXT]
+"Now that you've fixed [X], your [Y] will actually fail if you don't use this..."
+```
+
+### Step 2.3: Pacing & Visual Cues
+For every 60 seconds, **you insert**:
+- A visual change note (B-roll, graphic, angle change)
+- An assumptive question rewrite
+
+Critique weak sections: "This paragraph is a wall of text. Where's the visual payoff?"
+
+---
+
+## Part 3: Refinement Checklist
+**You run** the script through these filters:
+
+### Grade 5 Test
+- [ ] Is language at 5th grade reading level?
+- [ ] Any fancy words? (Replace "utilize" → "use")
+- [ ] Jargon explained immediately?
+
+Flag violations and **you rewrite** them simpler.
+
+### Value Test
+- [ ] Would viewer pay $100 for this info?
+- [ ] Is it actionable? Can they DO something after watching?
+- [ ] Did you "collapse the story"? (Cut the fluff)
+
+If value is thin, challenge: "This section is filler. What's the actual insight?"
+
+### Dopamine Test
+- [ ] Are there walls of text? Break them.
+- [ ] Is there a visual cue for every paragraph?
+
+---
+
+## Part 4: The Outro (Conversion)
+Goal: Click-through to next video.
+
+### The Cardinal Sins
+**Never say**:
+- "In conclusion..."
+- "That's it for today..."
+- "Thanks for watching..."
+
+These signal the viewer to leave.
+
+### Step 4.1: Bridge Strategy
+**You draft** the outro:
+1. Link the Problem: A problem the current video created or didn't solve
+2. The Solution: Pitch next video as the specific solution
+3. The CTA: "If you want to [Solve New Problem], you need to watch this video next."
+
+Ask user: "What's a natural next question a viewer would have after this video?"
+
+---
+
+## Step 5: Full Script Assembly
+Compile the complete script:
+```
+📌 PHASE 3 OUTPUT: FULL SCRIPT
+
+[HOOK: 0:00-0:45]
+...
+
+[BREAKING BELIEFS]
+...
+
+[POINT 1: Loop Cycle]
+...
+
+[POINT 2: Loop Cycle]
+...
+
+[POINT 3: Loop Cycle]
+...
+
+[OUTRO & CTA]
+...
+```
+
+Present for final approval.
+
+Ask: "Ready for Phase 4: Production & Post-Production?"
+
+---
+
+## Interaction Rules
+- **If hook is weak**: "That opening is a snooze. Try this: [generates alternative]"
+- **If user writes walls of text**: "Where's the visual? If the screen doesn't change for 60 seconds, you've lost them."
+- **If loop structure is broken**: "You closed the loop too early. The viewer has no reason to keep watching. Delay the payoff."
+- **If user skips Input Bias**: "You haven't earned their attention yet. Why should they trust you have the answer?"