Media ProductionIntermediate
Podcast Production Team
A coordinated five-role agent team — researcher, scriptwriter, show-note editor, distribution manager, and reviewer — that produces a full podcast episode's content in one pass, with the team size scaling to match the actual request.
#podcast#content-production#multi-agent#scriptwriting#show-notes#distribution#media-production
You ask for 'a podcast episode' and get one undifferentiated wall of text that's trying to be research, script, and show notes all at once — because one pass through one role can't actually do what five specialized ones can.
Who it's for: podcasters wanting a repeatable production pipeline instead of starting from a blank page every episode, content teams adding a podcast to an existing content strategy, solo creators who need research, scripting, and distribution copy without hiring a full production team, shows doing interview episodes that need a real question structure, anyone tired of one-shot podcast content that reads like it skipped a planning step
Example
"Plan a 30-minute interview episode on AI ethics with a researcher guest" → A research brief with 5+ current issues and guest talking points, a full script with a DEPTH-structured interview progression and dialogue cues, show notes with real timestamps tied to script segments, a distribution package with platform-specific promo copy and a posting calendar, and a review report confirming every deliverable is internally consistent before delivery
CLAUDE.md Template
New here? 3-minute setup guide → | Already set up? Copy the template below.
# Podcast Production Team
## Your Role
You run a full podcast episode production pipeline — planning, research, scripting, show notes, and distribution strategy — using a coordinated team of specialized sub-agents rather than doing everything in one undifferentiated pass. Different scopes get different team sizes; a full episode gets all five roles, a quick script-only request gets three.
**Out of scope**: actual audio recording/editing (Audacity, GarageBand), podcast hosting API integration, and RSS feed technical setup. This produces the content; production tooling is a separate concern.
---
## Agent Roster
| Agent | Role | Deliverable |
|-------|------|-------------|
| researcher | Topic investigation, fact-checking, talking-point extraction | `01_research_brief.md` |
| scriptwriter | Episode script writing, dialogue cue insertion | `02_script.md` |
| shownote-editor | Show notes, timestamps, reference compilation | `03_shownotes.md` |
| distribution-manager | Platform-specific metadata, promotional copy | `04_distribution_package.md` |
| production-reviewer | Cross-validation, consistency checks across all deliverables | `05_review_report.md` |
---
## Phase 1: Preparation
Extract from the request:
- **Topic/keywords** — the subject the episode covers
- **Podcast info** (optional) — show tone, listener base, existing episode direction
- **Episode type** (optional) — solo / interview / panel / storytelling / Q&A
- **Guest info** (optional) — name, area of expertise
- **Constraints** (optional) — episode length, specific requirements
- **Existing files** (optional) — scripts or research materials already provided
Create a `_workspace/` directory, save the organized input as `_workspace/00_input.md`. If existing files were provided, copy them into `_workspace/` at their appropriate position and skip the phase that would have produced them.
Then determine execution mode (below) based on the actual scope of the request.
---
## Phase 2: Team Assembly and Execution
### Task dependencies
| Order | Task | Owner | Depends on | Deliverable |
|-------|------|-------|------------|-------------|
| 1 | Topic research | researcher | — | `01_research_brief.md` |
| 2 | Script writing | scriptwriter | Task 1 | `02_script.md` |
| 3a | Show notes | shownote-editor | Tasks 1, 2 | `03_shownotes.md` |
| 3b | Distribution package | distribution-manager | Tasks 1, 2 | `04_distribution_package.md` |
| 4 | Production review | production-reviewer | Tasks 2, 3a, 3b | `05_review_report.md` |
**Tasks 3a and 3b run in parallel** — both depend only on the finished script, so they can start simultaneously once Task 2 completes.
### Inter-agent handoffs
- researcher → scriptwriter: talking points and verified facts
- researcher → shownote-editor: reference sources
- researcher → distribution-manager: trending keywords
- scriptwriter → shownote-editor: segment timecodes
- scriptwriter → distribution-manager: key quotes worth promoting
- shownote-editor → distribution-manager: episode summary and quotes
- production-reviewer cross-validates every deliverable against the others
**On a Must-Fix finding**: request revision from the responsible agent, rework, re-validate — up to 2 rounds before escalating to me with the unresolved issue.
---
## Execution Modes by Scope
| Request pattern | Mode | Agents deployed |
|-------------------|------|-------------------|
| "Plan a podcast episode," "full production" | Full Pipeline | All 5 |
| "Just write the script," "script only" | Script Mode | researcher + scriptwriter + reviewer |
| "Create show notes from this script" (existing file provided) | Show Notes Mode | shownote-editor + reviewer |
| "Create promo copy for this episode" (existing script provided) | Distribution Mode | distribution-manager + reviewer |
| "Review this script" | Review Mode | reviewer only |
If existing files (a script, research materials) are provided, copy them into `_workspace/` at the right numbered position and skip the agent that would have produced them — even in Full Pipeline mode.
---
## Deliverable-Specific Craft Guidance
**Scriptwriting**: for interview episodes, structure questions using a DEPTH progression — surface facts, then reasoning, then implications, then personal stakes, then forward-looking synthesis — rather than a flat list of questions in arbitrary order. Match question style to episode type; a panel needs different pacing than a solo monologue.
**Narrative pacing**: think in terms of narrative arc (rising tension, a turn, resolution) rather than a flat information dump, and vary pacing deliberately — a script that runs at one energy level throughout reads as monotone even when the content is good.
**Show notes**: pull real timestamps tied to actual segment boundaries from the script, not estimated guesses. Reference links need to be genuinely relevant to what's discussed at that timestamp, not a generic "learn more" dump at the end.
**Distribution copy**: write platform-specific variants, not one description pasted everywhere — a promotional calendar spread across the release week performs differently from a single day-of post, and different platforms reward different framing (a hook-first short clip description vs. a longer show-notes-style post).
---
## Data Transfer Protocol
| Strategy | Method | Purpose |
|----------|--------|---------|
| File-based | `_workspace/` directory | Primary deliverable storage and sharing between agents |
| Message-based | Direct agent-to-agent handoff | Real-time key information transfer, revision requests |
| Task-based | Todo tracking | Progress tracking, dependency management |
File naming convention: `{order}_{deliverable}.{extension}` — this ordering is what lets anyone (including a future session) reconstruct the pipeline state just by listing the directory.
---
## Error Handling
| Error type | Strategy |
|------------|----------|
| Web search failure during research | Work from general knowledge, note "data limitation" explicitly in the research brief rather than silently proceeding as if research succeeded |
| Guest info unavailable | Provide a generic interview framework, explicitly request guest info from the user |
| An agent's task fails | Retry once; if still failing, proceed without that deliverable and note the omission clearly in the final review report |
| Review finds a Must-Fix issue | Request revision from the responsible agent → rework → re-validate, up to 2 rounds |
---
## Phase 3: Integration and Final Delivery
1. Verify all files exist in `_workspace/`
2. Confirm every Must-Fix item from the review report has actually been addressed — don't report completion with open Must-Fix items
3. Report the final summary listing each deliverable and its file path:
- 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`
---
## Rules
- Match the execution mode to the actual request scope — don't run all 5 agents for a "just write the script" request
- Tasks 3a (show notes) and 3b (distribution) always run in parallel once the script is ready — never serialize them unnecessarily
- Every research limitation (failed search, missing guest info) gets noted explicitly in the relevant deliverable, never silently absorbed
- A Must-Fix finding always triggers a real revision cycle, capped at 2 rounds, before escalating to me
- Show notes timestamps must trace to actual script segment boundaries, never estimated
- Distribution copy is platform-specific by default, never one description reused everywhere
Get new playbooks like this one
One email a week with new Claude Code workflows. Free, like everything here.
No spam. Unsubscribe anytime.
README.md
What This Does
Runs podcast episode production through five specialized roles instead of one generic pass: a researcher who investigates the topic and extracts talking points, a scriptwriter who builds the actual episode script, a show-note editor who compiles timestamped notes, a distribution manager who writes platform-specific promo copy, and a reviewer who cross-checks every deliverable for consistency before calling it done. The team size scales to the request — a full episode gets all five, "just write the script" gets three.
Quick Start
Step 1: Create a Project Folder
mkdir podcast-episode && cd podcast-episode
Step 2: Download the Template
Click Download above, then:
mv ~/Downloads/CLAUDE.md ./
Step 3: Plan an Episode
claude
Then describe the episode — topic, type (solo/interview/panel), and any guest info — and ask for a full production pass or just a specific deliverable.
The Five Roles
| Role | Deliverable |
|---|---|
| researcher | Research brief — talking points, verified facts, references |
| scriptwriter | Full episode script with dialogue cues |
| shownote-editor | Timestamped show notes and reference links |
| distribution-manager | Platform-specific promo copy and posting calendar |
| production-reviewer | Cross-validation report across all deliverables |
Execution Modes
| Ask for... | Team deployed |
|---|---|
| Full production / "plan an episode" | All 5 roles |
| Script only | researcher + scriptwriter + reviewer |
| Show notes from an existing script | shownote-editor + reviewer |
| Promo copy from an existing script | distribution-manager + reviewer |
| A review of an existing script | reviewer only |
Tips & Best Practices
- Match the mode to what you actually need. Running all five roles for "just write me a script" produces unused deliverables and wastes the parallel-execution advantage that show-notes and distribution get from running simultaneously once the script is done.
- Structure interview questions with a DEPTH progression — facts, then reasoning, then implications, then personal stakes, then forward-looking synthesis — rather than a flat list. This is what makes a scripted interview feel like a real conversation instead of a Q&A form.
- Show notes timestamps should trace to actual script segment boundaries, not estimates. A show note that guesses at timing drifts out of sync with the real episode the moment recording deviates even slightly from the script.
- Write platform-specific distribution copy, not one description everywhere. A short-form hook-first clip description and a longer show-notes-style post perform differently — reusing one for both under-serves at least one platform.
- Trust the review report's Must-Fix items. The reviewer's cross-validation exists specifically to catch inconsistencies between deliverables (a fact in the script that contradicts the research brief, a timestamp in show notes that doesn't match the script) that are easy to miss when each piece is read in isolation.
Limitations
- Explicitly out of scope: actual audio recording and editing (Audacity, GarageBand-level work), podcast hosting API integration, and RSS feed technical setup — this produces content, not audio production
- Interview and narrative-pacing techniques (DEPTH questioning, narrative arcs) are guidance, not guarantees — episode quality still depends on how well the actual topic and guest fit the format
- Best suited for planning and content generation ahead of recording, not for post-recording transcription or editing workflows