Podcast Production Skill

Overview

Transform a raw podcast transcript into polished, multi-platform content assets through four strategic checkpoints. Each checkpoint delivers decision-ready analysis in a markdown file for your feedback before proceeding. You’ll provide feedback directly in the checkpoint documents, then we’ll iterate before moving to the next phase. Final outputs include publication-ready YouTube strategy and a narrative-driven blog post.

Workflow Structure:

• Start with [Guest]_Source_Material.md (raw transcript + notes)
• Create Checkpoint_1_Comprehensive_Analysis.md (your feedback here)
• Create Checkpoint_2_Cold_Opens_and_Clips.md (your feedback here)
• Create Checkpoint_3_YouTube_Strategy.md (your feedback here)
• Create Checkpoint_4_Polished_Transcript_and_Blog.md (final deliverable)

Execution Model: Subagents for Context Preservation

CRITICAL: Each checkpoint should be executed by a dedicated subagent (using the Task tool) to preserve context window in the main conversation. The workflow is:

1. Main agent reads source material, confirms approach with user
2. Subagent 1 executes Checkpoint 1, writes file, reports back summary
3. User reviews, provides feedback on Big Idea selection
4. Subagent 2 executes Checkpoint 2 (reads Checkpoint 1 + source), writes file, reports back
5. User reviews, approves cold open and clips
6. Subagent 3 executes Checkpoint 3 (reads prior checkpoints), writes file, reports back
7. User reviews, approves YouTube strategy
8. Subagent 4 executes Checkpoint 4 (reads all prior work), writes final deliverables

Subagent prompt template:

You are executing [Checkpoint N] of the podcast production workflow.

Episode: [Guest Name]
Working directory: [path to episode folder]

Read the following files:
- SOURCE.MD (raw transcript)
- [Any prior checkpoint files]

Create: Checkpoint_[N]_[Name].md following the podcast-production skill format.

[Specific checkpoint instructions from skill]

Write the checkpoint file and report back with:
1. Key decisions/recommendations
2. Questions for user feedback

This preserves context by having each subagent start fresh with only the necessary files, rather than accumulating the entire conversation history.

When to Use This Skill

• You have a raw podcast transcript and need to identify the strongest marketing angle
• You want to create a cold open that hooks listeners immediately
• You need YouTube titles and thumbnail strategies
• You’re creating social clips from podcast material
• You want to create SEO-optimized blog content from the episode
• You want all assets aligned with OpenEd brand identity

Session Startup

At the start of each podcast production session, sync the content database:

cd "Content/Misc. Utilities/seomachine" && python3 -c "
from dotenv import load_dotenv
load_dotenv('data_sources/config/.env')
from data_sources.modules.webflow import sync_content_database
sync_content_database()
"

This pulls latest content from Webflow to Content/Master Content Database/ for internal linking and context.

THE FOUR CHECKPOINTS

Checkpoint 1: Comprehensive Analysis (90-120 min)

Goal: Understand the episode, identify strongest themes, and inventory building blocks for all downstream content (social clips, long-form clips, blog posts, newsletters).

Your deliverable: Checkpoint_1_Comprehensive_Analysis.md

Document Structure (in this order):

1. Episode Metadata (guest, host, duration, credentials)
2. The TED Talk Version (2-3 orthogonal angles that capture what the episode delivers – emphasize salience over clickability)
3. 5 Big Ideas (potential marketing angles for different assets)
4. Chapter Outline (timestamped)
5. Snippet Inventory (25-30 short quotable moments, 5-30 sec each, labeled S1-S29)
6. Guest Voice Inventory (sticky phrases, syntax patterns, named concepts)
7. Quote Bank (organized by theme)
8. Surprising Points (contradictions with common belief)
9. Top Recommendation (with reasoning)

Snippet Inventory Format – Each snippet MUST include:

• Label (S1, S2, etc.)
• Category (Counterintuitive, Memorable Quotes, Relatable/Funny, Practical/Actionable, Emotional, Origin Story)
• Timestamp range
• Speaker attribution (JOSHUA, ISAAC, ELA, etc.)
• Verbatim quote

Example:

**[S1] "No Shoulds, Only Coulds" (18:21-18:52) - JOSHUA**
> "There are no shoulds, there are only coulds..."

The TED Talk Version – Provide 2-3 somewhat orthogonal angles:

• What is this episode actually about? (salience)
• What’s the surprising/counterintuitive framing? (interest)
• What’s the practical takeaway? (utility)

User decision point: Which Big Idea/angle to pursue for primary assets?

IMPORTANT: Balance clickability with salience. Lean toward what the episode actually delivers rather than what might get the most clicks but misrepresent the content.

Reference: See references/checkpoint-1-example.md for detailed example

Checkpoint 2: Cold Opens & Clips (120-150 min)

Prerequisites: Checkpoint 1 complete + Big Idea selected

Goal: Create one approved cold open script and identify 3-5 approved social clips with on-screen hooks.

Your deliverable: Checkpoint_2_Cold_Opens_and_Clips.md

Process (two-step):

1. First: Arrange snippets into cold open options — Using the Snippet Inventory from Checkpoint 1, combine snippets (by label: S1, S2, etc.) into 3-5 cold open arrangements. Use [SWOOSH] to indicate transitions between unrelated moments.
2. Then: Expand selected snippets into full clips — Build out the best snippets into complete 45-90 second clips with full verbatim transcripts and edit markup.

What it contains:

• 3-5 cold open options (montages of snippets with [SWOOSH] transitions)
• One selected cold open script (22-35 seconds)
• On-screen text hooks (2-4 words each for social media captions)
• 3-5 approved social clips (45-90 sec each, full verbatim transcripts with edit markup)
• Each clip includes: timestamp range, on-screen hook, full transcript with strikethrough for cuts and italics for minor edits

Output format:

• Cold open: Montage structure using [SWOOSH] between unrelated moments, not linear narrative extracts
• Clips: Verbatim with edit markup (cuts, changes), timestamps throughout
• Hooks: 2-4 words maximum for mobile readability
• All clips ready for editor handoff

Skills used:

• video-caption-creation: For on-screen text hooks and short-form video captions (generates 3-5 hook options per clip)
• cold-open-creator: For cold open methodology (optional reference)

Clip Types (create a mix):

Supercut Template:

**On-screen hook options:**
- "[Caption option 1]"
- "[Caption option 2]"

**Format:** Quick cuts, [SWOOSH] between each moment. ~45-60 seconds total.

---

**SPEAKER (timestamp):** [Short moment - 5-10 sec]

[SWOOSH]

**SPEAKER (timestamp):** [Short moment - 5-10 sec]

[SWOOSH]

(continue for 5-8 moments, end on emotional or memorable beat)

User decision point: Approve cold open and social clips for Checkpoint 3

Reference: See references/checkpoint-2-example.md for detailed example

Checkpoint 3: YouTube Strategy (90-120 min)

Prerequisites: Checkpoint 2 complete + Cold open selected

Goal: Define YouTube title, thumbnail, description, and chapter timestamps.

Your deliverable: Checkpoint_3_YouTube_Strategy.md

What it contains:

• Final YouTube title (with guest name for authority)
• Thumbnail specification (2-4 words max, minimal design)
• YouTube description (opening hook sentence + full description + resources + chapters)
• Cold open script (verbatim clips)
• All 3 approved social clips with on-screen hooks
• Chapter breakdown (5-10 words per chapter title, keyword-rich)

Skills used:

• youtube-title-creator: For YouTube title strategy
• opened-identity: For brand alignment verification

Output format:

• Clean, streamlined specifications (no technical jargon)
• Chapter titles follow “My First Million” style (compelling, descriptive)
• Format: (MM:SS) - Descriptive Chapter Title (5-10 words max)
• Description: Opening hook + full description + resources + chapters
• Thumbnail: Simple visual + minimal text (2-4 words only)
• All clips ready to copy/paste

User decision point: Approve final specifications and move to Checkpoint 4

Reference: See references/checkpoint-3-example.md for detailed example

Checkpoint 4: Polished Transcript & Blog Post (120-180 min)

Prerequisites: Checkpoint 3 complete + All selections locked

Goal: Create publication-ready transcript and SEO-optimized blog post.

Your deliverables:

• [Guest]_YouTube_and_Show_Notes.md (refined from Checkpoint 3)
• [Guest]_Polished_Transcript.md (new, contains transcript + blog)

What it contains:

• Full polished transcript (cleaned for readability)
• Embedded blog post (~1,000 words) focused on core insight
• Guest bio and resource links
• SEO headers and structure

Skills used:

• transcript-polisher: For transcript cleanup and formatting
• podcast-blog-post-creator: For narrative-driven blog post creation in Ela’s voice

Output structure:

1. [Guest]_YouTube_and_Show_Notes.md — Handoff file for video production (title, thumbnail, cold open, show notes, timestamps)
2. [Guest]_Polished_Transcript.md — Publication-ready transcript with embedded blog post

Reference: See references/checkpoint-4-example.md for detailed example

WORKFLOW TIMELINE & FILE STRUCTURE

KEY PRINCIPLES

Verbatim Only

All quoted transcript must be exactly as spoken. You can cut/rearrange, never paraphrase.

Mine the Entire Transcript

Don’t limit analysis to one section. The strongest angle might be anywhere.

Bold Over Safe

Surprising, contrarian moments beat safe, obvious observations.

Story Over Summary

Create narrative momentum. Clips should have complete arcs, not just be “good quotes.”

Simple Over Complex

• Thumbnails: 3 elements max
• Titles: One clear idea
• Clips: Clear beginning, middle, end

Brand Aligned

All outputs reflect OpenEd visual and tonal guidelines. Homeschool parents should see themselves.

SKILL DEPENDENCIES

Optional skill triggers:

• day-in-the-life: When guest describes daily/weekly homeschool structure. Use Guest Voice Inventory to preserve their voice.
• verified-review: When guest mentions specific curricula/tools. Creates 300-500 word reviews with real parent attribution.

QUALITY GATES

COMMON MISTAKES

❌ Mining only part of transcript | ❌ Paraphrasing (all quotes verbatim) | ❌ Complex thumbnails (3 elements max) | ❌ Safe over surprising | ❌ Cold open too long (25-35s max) | ❌ Resolving cliffhanger | ❌ Skipping quality gates | ❌ Blog too formal (write like Ela)

REFERENCES

For detailed instructions and examples, see:

• references/checkpoint-1-example.md — Complete Checkpoint 1 example
• references/checkpoint-2-example.md — Complete Checkpoint 2 example
• references/checkpoint-3-example.md — Complete Checkpoint 3 example (YouTube specifications format)
• references/checkpoint-4-example.md — Complete Checkpoint 4 example (Polished transcript + blog post)

RELATED SKILLS

cold-open-creator | video-caption-creation | youtube-title-creator | podcast-blog-post-creator | transcript-polisher | opened-identity | day-in-the-life | verified-review

SUCCESS METRICS