SKILL.md
podcast-and-audiograms
The audio→social pipeline — clip the best moments (retention-led), brand + caption them, CTA back to the full episode. The agent plans + briefs + writes hooks/CTAs + plans the cadence; a clip/audiogram tool produces the clips; WoopSocial publishes the finished clips/audiograms on a schedule; the full episode lives on the podcast host.
The POV: a podcast is long-form; social is short, visual, sound-off
The skill isn't "make a podcast" — it's turning the best 30–90s moments into branded, captioned clips/audiograms that drive people back to the full episode (the episode is the pizza; a clip is one irresistible slice). The top-1% lever: select by listener-retention spikes (Spotify/YouTube Studio = replayed moments) + the timestamped transcript, not just an AI virality score — and a strategic CTA that gives a reason to watch the full episode. Audiograms (cover art + waveform + captions) make audio-only shows visual. The tools clip/generate; WoopSocial publishes the finished files; the podcast is hosted elsewhere.
Read these first
- brand-profile — voice/branding.
- the episode — recording + transcript + retention data.
The framework: WAVES
(Depth: references/the-waves-framework.md.)
- W — Winnow to the best moments: retention spikes (Spotify/YouTube Studio) + the timestamped transcript →
5–8 moments/episode; don't re-listen; AI virality scores are a starting point.
- A — Aim the format to what you recorded: video-podcast clip (9:16, reframe to active speaker) /
audiogram (cover art + waveform + captions) if audio-only / quote card (→ design-and-templates).
- V — Voice the hook + captions: first-3-sec hook; captions on every clip (85% sound-off — non-optional);
video-clip burn-in → captions-and-clipping/clipping.md.
- E — Endcap with a reason to listen: strategic CTA (a teaser/cliffhanger, not "check out the full
episode") + link to the episode on the host; SEO keywords.
- S — Spread on a schedule: WoopSocial publishes the finished clips/audiograms across platforms on a
cadence (episode → clips → quote cards → carousel); one episode → a week+ (→ content-recycling). The full episode stays on the host.
The reality (verify-quarterly)
55% of Americans listen monthly (Edison 2025); 85% watch sound-off → captions non-optional. Two formats: video-podcast clip (9:16, active-speaker reframe) + audiogram (cover art + animated waveform + captions) for audio-only — video clips outperform audiograms (attribute), so prefer video where it exists and build audiograms caption-first (waveform as support, not the star); quote cards for a line. Pipeline: transcribe (Descript/Castmagic) → timestamp standout moments + retention spikes → 5–8 clips → hook + captions + branding → strategic CTA → publish. Tools: Opus Clip (virality score, reframe; not an editor — little creative control), Choppity (best conversational accuracy), Descript (the editor), Kapwing (free); audiograms: Headliner/Podsqueeze; written: Castmagic; design: Canva (→ design-and-templates): references/podcast-to-social-2026-reality.md. The production pipeline, format decision + specs, the tool table, the audiogram anatomy, the CTA library, the WoopSocial flow + worked examples: references/pipeline-formats-and-tools.md.
Honest scope (never violate)
- The agent plans the clip strategy (retention-based), briefs the clip/audiogram creation, **writes
hooks + CTAs, and plans the cadence.**
- A clip/audiogram tool produces the clips; the creator finalizes them; **WoopSocial publishes the
finished clips/audiograms on a schedule. WoopSocial does NOT record/edit/transcribe the podcast, generate clips/audiograms, host the podcast (Spotify/Apple/RSS), or pull retention analytics.**
- Captions non-optional (85% sound-off); strategic CTA to the full episode; **the full episode lives on
the host; never fabricate virality; never clip someone else's copyrighted podcast** for a brand. (Scope, distinctions + connections: references/scope-and-connections.md.)
Distinct from its siblings (route correctly)
podcast-and-audiograms (this) = the audio→social pipeline (retention-led selection + the audiogram format + CTA-back-to-episode) · captions-and-clipping = the general clip/caption-burn-in craft (a video-podcast clip's burn-in routes there) · content-recycling/cross-platform-repurposing = the general reuse axes (the episode is one source) · youtube-long-form/youtube-publishing-and-metadata = the full episode on YT · ai-voiceover = a synthetic voice (this repurposes a real recording) · ai-music-and-sound = the music bed · carousel-writer/email-and-newsletter = the quote-card + newsletter routes.
Where this connects
Reads first: brand-profile + the episode (recording + transcript + retention). Pulls from: the podcast host (the full episode + retention), the clip/audiogram tools, captions-and-clipping/ clipping.md (burn-in), design-and-templates (audiogram templates + quote cards), carousel-writer (a moment → carousel), email-and-newsletter (episode → newsletter), youtube-publishing-and-metadata (titles/keywords). Publishes via: the finished clips/audiograms → scheduling-and-queue → WoopSocial (cross-platform cadence). Reuse: content-recycling/cross-platform-repurposing. Measure with: native + analytics-and-reporting (never fabricated). The full episode stays on the podcast host.
Definition of done
An episode repurposed into a week of social: 5–8 moments chosen from retention spikes + the timestamped transcript; the right format per moment (9:16 video clip with active-speaker reframe / audiogram with cover art + waveform + captions / quote card); a first-3-second hook and captions on every asset (non-optional); a strategic CTA that gives a reason to watch the full episode, linked to the host; an on-brand template; the clips produced by a clip/audiogram tool and finalized by the creator, then published by WoopSocial across platforms on a cadence; recording/editing/hosting/transcription left with the tools + the podcast host; no virality fabricated and no one else's copyrighted podcast clipped; correctly distinguished from captions-and-clipping, content-recycling, and youtube-long-form.