Podcast Studio

Podcast Studio — Full Podcast Production Pipeline

An agent team collaborates to produce podcast episodes through the pipeline of planning → research → scripting → show notes → distribution, all at once.

Execution Mode

Agent Team — 5 members communicate directly via SendMessage and cross-validate each other's work.

Agent Roster

Workflow

Phase 1: Preparation (Performed directly by the orchestrator)

1. Extract from user input:
   • Topic/Keywords: The subject the episode will cover
   • Podcast Info (optional): Show tone, listener base, existing episode direction
   • Episode Type (optional): Solo/Interview/Panel/Storytelling/Q&A
   • Guest Info (optional): Guest name, area of expertise
   • Constraints (optional): Episode length, specific requirements
   • Existing Files (optional): Scripts, research materials, etc. provided by the user
2. Create the _workspace/ directory at the project root
3. Organize the input and save it as _workspace/00_input.md
4. If existing files are provided, copy them to _workspace/ and skip the corresponding phase
5. Determine the execution mode based on request scope (see "Scope-Based Modes" below)

Phase 2: Team Assembly and Execution

Assemble the team and assign tasks. Task dependencies are as follows:

Tasks 3a (show notes) and 3b (distribution) run in parallel. Both depend on Task 2 (script), so they can start simultaneously once the script is complete.

Inter-agent communication flow:

• researcher complete → deliver talking points & facts to scriptwriter, references to shownote-editor, trending keywords to distribution-manager
• scriptwriter complete → deliver segment timecodes to shownote-editor, key quotes to distribution-manager
• shownote-editor complete → deliver episode summary & quotes to distribution-manager
• reviewer cross-validates all deliverables. On 🔴 Must Fix: request revision from the responsible agent → rework → re-validate (up to 2 rounds)

Phase 3: Integration and Final Deliverables

Compile the final deliverables based on the reviewer's report:

1. Verify all files in _workspace/
2. Confirm all 🔴 Must Fix items from the review report have been addressed
3. Report the final summary to the user:
   • Research Brief — 01_research_brief.md
   • Episode Script — 02_script.md
   • Show Notes — 03_shownotes.md
   • Distribution Package — 04_distribution_package.md
   • Review Report — 05_review_report.md

Scope-Based Modes

Adjust the agents deployed based on the scope of the user's request:

Using Existing Files: If the user provides a script, research materials, or other existing files, copy them to the appropriate numbered position in _workspace/ and skip that phase's agent.

Data Transfer Protocol

File naming convention: {order}_{deliverable}.{extension}

Error Handling

Test Scenarios

Happy Path

Prompt: "Plan a 30-minute interview podcast episode on AI ethics. The guest is an AI researcher" Expected Results:

• Research: 5+ latest AI ethics issues, guest background, 5 talking points
• Script: 30-minute length (~4,500 words in English), opening + 3–4 segments + closing, guest questions included
• Show notes: Timestamps, reference links, key quotes
• Distribution: Platform-specific descriptions, 3 social media copy variants, promotional calendar
• Review: All consistency matrix items verified

Existing File Flow

Prompt: "Create show notes and a distribution package from this script" + script file attached Expected Results:

• Existing script copied to _workspace/02_script.md
• Show Notes Mode + Distribution Mode merged: shownote-editor + distribution-manager + reviewer deployed
• researcher, scriptwriter skipped

Error Flow

Prompt: "Just write a podcast script quickly, any topic" Expected Results:

• Switch to Script Mode (researcher + scriptwriter + reviewer)
• Topic is unclear, so researcher proposes 3 trending topics before proceeding
• Review report notes "show notes/distribution package not generated"

Agent Extension Skills

Each agent leverages the following extension skills to enhance deliverable quality: