recoup-release-pack

What this skill does

# Release Pack

A pre-release marketing brief. The skill exists because generic AI-generated
marketing copy is slop, and an LLM asked for "marketing ideas" with no
grounding will always produce slop. This skill produces **structured input**
that's grounded in real audience data and the user's stated creative direction
— not finished marketing assets.

The output is the scaffold: visualizer directions, content angles, playlist
targets, narrative thread. The user renders those into actual videos, copy,
and creative through whatever pipeline they have.

## When to use

- "Release pack for {release name}"
- "Prep {release}" / "Rollout plan for {release}"
- "Marketing for {release}" / "What should I do for {release}"
- "Pre-release brief for {release}" / "Marketing pack for {release}"

For ongoing weekly artist tracking, use `recoup-weekly-brief`. For one-shot
artist research with no release context, use `recoup-artist-research`.

## Setup

```bash
export RECOUP_API_KEY="recoup_sk_..."
export RECOUP_API="https://api.recoupable.com/api"
```

## What you need from the user (refuse to proceed without it)

A release pack requires **release-specific facts** the API does not have. Before
fetching anything, confirm with the user:

1. **Artist** (workspace slug or provider artist ID)
2. **Release name** (track or album title)
3. **Release date** (or "TBD" if not set — still useful)
4. **Release type** (single, EP, album, deluxe, remix, collab, soundtrack)
5. **One-line creative direction** — vibe / themes / inspirations / collaborators
6. **What asset gaps exist** — e.g., "no music video yet", "no cover art lock",
   "we have stems but no remix plan"

If any of #2, #3, #5 are missing, **ask** — don't invent. A release pack about
a release the agent has no concrete information about is exactly the slop this
skill is meant to prevent.

## Workflow

### 1. Confirm release facts (above)

Ask once, get the 6 inputs, proceed. Do not loop on this.

### 2. Pull artist context (in parallel)

```bash
# Audience demographics — drives platform mix
curl -s "$RECOUP_API/research/audience?artist={ARTIST}&platform=tiktok" \
  -H "x-api-key: $RECOUP_API_KEY" &

curl -s "$RECOUP_API/research/audience?artist={ARTIST}&platform=instagram" \
  -H "x-api-key: $RECOUP_API_KEY" &

# YouTube audience — extra geographic/demo signal
curl -s "$RECOUP_API/research/audience?artist={ARTIST}&platform=youtube" \
  -H "x-api-key: $RECOUP_API_KEY" &

# Similar artists — drives playlist targets + collab/feature ideas
curl -s "$RECOUP_API/research/similar?artist={ARTIST}&audience=high&genre=high&limit=20" \
  -H "x-api-key: $RECOUP_API_KEY" &

# Current playlist position — drives gap analysis for new pitches
curl -s "$RECOUP_API/research/playlists?artist={ARTIST}&status=current" \
  -H "x-api-key: $RECOUP_API_KEY" &

wait
```

Geography comes from the country breakdown in `/research/audience` (the
`cities` endpoint was removed). For venue/festival/touring detail, add a
`POST /research/web` query.

### 3. (Optional) Track lookup if the release is already in the provider

If the release is already ingested, resolve its provider track ID and pull
detail:

```bash
curl -s "$RECOUP_API/research?q={RELEASE_NAME}&type=tracks" \
  -H "x-api-key: $RECOUP_API_KEY"                              # get track id
curl -s "$RECOUP_API/research/track?id={track_id}" \
  -H "x-api-key: $RECOUP_API_KEY"
```

For most pre-release packs, the track isn't ingested yet — skip this step.

### 4. Synthesize the pack

Write to `artists/{slug}/releases/{release-slug}/pack-$(date +%F).md`.
The `{release-slug}` is the kebab-case release title.

Template:

```markdown
# Release Pack — {Artist} — *{Release Name}*
**Type:** {single|EP|album|...} | **Release date:** {YYYY-MM-DD or "TBD"} | **Generated:** {YYYY-MM-DD}

## Creative direction (from user)
{The 1-line creative direction the user gave verbatim. This is the source of truth.
Everything below is in service of THIS direction.}

## Narrative thread
{2–4 sentences. The cohesive story for the rollout. What's the one thing
listeners should remember after the campaign. Anchored in the creative direction,
not invented.}

## 3 visualizer concept directions

### Direction A — {short name, e.g., "Mood Piece"}
- **Concept (2 sentences):** ...
- **Visual references:** {3–5 reference touchpoints — artists, films, eras}
- **Production bar:** {low / med / high — be honest about what this needs}
- **Why this for {Artist}:** {tie to audience data — TT skew, geo, similar-artist visual language}

### Direction B — {short name, e.g., "Narrative"}
{Same structure}

### Direction C — {short name, e.g., "Archive / Lo-Fi"}
{Same structure}

## 5 content angles

Each angle is a *premise for a content series*, not a single post. Each should
be sustainable across 4–8 pieces of content.

| # | Angle | Why it works for THIS release | First 3 piece ideas |
|---|---|---|---|
| 1 | {name} | {tie to audience + creative direction} | {3 concrete posts} |
| 2 | ... | ... | ... |
| ... | ... | ... | ... |

## Platform hooks

| Platform | Audience signal | Hook approach |
|---|---|---|
| TikTok | {top age/gender/geo from audience endpoint} | {1–2 hook formats matched to that audience} |
| IG Reels | {top age/gender/geo from audience endpoint} | {1–2 hook formats} |
| YouTube Shorts | {if data available} | {hook format} |
| Twitter/X | {if relevant for this artist's lane} | {hook format} |

## Playlist targets (ranked)

| Playlist | Followers | Why this fits | Warm path? |
|---|---|---|---|
| {top from similar artists' placements that this artist isn't on} | {followers_total} | ... | {yes/no/unknown} |
| ... | ... | ... | ... |

Source the targets from `/research/playlists` cross-referenced against
`/research/similar` — playlists where peers sit and this artist doesn't.

## Geographic rollout

Top markets for the artist (country breakdown from `/research/audience`, cross-
platform):
1. {country, % of audience, which platform}
2. ...
3. ...

Rollout implications: {1–2 sentences on which markets to weight, e.g., "Geo-pinned
content for the top 3 countries; physical events viable in the strongest market
if budget allows." For city-level or venue detail, run a `/research/web` query.}

## Open creative decisions (for the user to resolve)

- {gap from input #6, or surfaced from analysis}
- ...

## What this pack is NOT

This is structured input for downstream creative work. It is not:
- Final copy
- Finished video scripts
- Approved playlist pitch emails (use `recoup-playlist-intelligence` for that)
- Cover art or visualizer assets

Take the directions above into whatever creative pipeline you use, and produce
the assets.

---
*Generated {ISO timestamp}. Re-run with `/recoup-release-pack {release-slug}`.*
```

### 5. Print a short chat summary

3–5 sentences: the creative direction restated, the 3 visualizer direction
names, and the top 3 content angle names. So the user can decide whether to
open the file.

## What this skill refuses to do

- **No inventing the release.** If the user can't tell us what the release
  sounds like or what it's about in one line, this skill stops. A release pack
  for a release the agent has no creative knowledge of will produce slop —
  refuse instead.
- **No fake playlist names.** Playlist targets come from `/research/playlists`
  + `/research/similar`. If no good targets surface, the section says so.
- **No copy generation.** This is a brief, not assets.
- **No hallucinated audience numbers.** If `/research/audience` returns thin
  data for the platform, the platform hooks section says "thin audience data —
  weight by similar-artist behavior instead."

## Audience-driven hooks

The platform hooks section is the most-likely-to-go-slop part. To prevent that,
anchor every hook in a concrete audience data point:

❌ "TikTok hook: relatable Gen Z humor about heartbreak"
✅ "TikTok hook: top audience is F 18–24 in NYC + LA (37% of artist's TT
   audience). Hook format that's worked for simila