process-album

Delegation Rule

This skill is a pure orchestrator. It MUST invoke other skills using the Skill tool:

Skill("media-management:<skill-name>", args="<arguments>")

DO NOT call underlying scripts directly via Bash. DO NOT re-implement sub-skill logic inline. When a step says → Invoke:, use the Skill tool exactly as shown.

The ONLY exception is Step 2 and Step 11 (ZIP extraction), which use this skill's own script.

Setup

1. Check environment variables: MEDIA_MGMT_DOWNLOADS, MEDIA_MGMT_LIBRARY_IMPORT, MEDIA_MGMT_LIBRARY_STORAGE, MEDIA_MGMT_ARCHIVE_WORKDIR
2. For any unset variables, use the default paths from CLAUDE.md
3. If CLAUDE.md has no paths, read config.json from the project root

Scripts are in scripts/ relative to this skill directory.

Scripts

• extract-zip.sh <zip_file> <dest_folder> — Extract a ZIP into a named subfolder with safety checks. Run --help for details.

References

See references/safety-rules.md for critical safety rules. See references/album-types.md for album classification.

Workflow

Phase 1: MP3 Processing

Step 1: Identify release → Invoke: Skill("media-management:select-release")

• If a release is specified in $ARGUMENTS, pass it as args
• NEVER trust filename suffixes — the skill inspects archive contents

Step 2: Extract MP3 ZIP

Run the extraction script:

bash scripts/extract-zip.sh "$MP3_ZIP" "$DOWNLOADS/$RELEASE_NAME"

The script creates the subfolder, extracts the ZIP, and outputs JSON with the file list. If extraction fails, ask user to extract manually via Finder.

Step 3: Inspect metadata → Invoke: Skill("media-management:manage-metadata", args="inspect $EXTRACTION_FOLDER")

• Review the output for missing/inconsistent fields

Step 4: MANDATORY metadata verification — STOP AND ASK USER

• Present: track count, detected artist, album title
• Present numbered genre list from the inspect results, ask user to select or type custom
• Ask user to verify artist name
• Ask user to verify album title
• Check for multiple artists — if found, ask: "Is this a compilation?"
• See references/album-types.md for classification
• DO NOT PROCEED without user confirmation of ALL fields

Step 5: Check for long tracks

• Check duration of each MP3 (available from the inspect output)
• If any track > 78 minutes (4680 seconds), inform user and ask about splitting → If user confirms: Skill("media-management:split-long-tracks", args="$EXTRACTION_FOLDER")
• After splitting, update track count via manage-metadata

Step 6: Update metadata → Invoke: Skill("media-management:manage-metadata", args="update $EXTRACTION_FOLDER genre=$GENRE clear-album-artist|set-album-artist=$ARTIST update-track-count")

• Pass user-confirmed values (genre, album artist, compilation flag)

Step 7: Import to Apple Music → Invoke: Skill("media-management:import-to-apple-music", args="$EXTRACTION_FOLDER")

Step 8: MANDATORY Apple Music verification — STOP AND ASK USER

• Tell user: "Please check Apple Music and confirm the import looks correct"
• Check for: files appearing as a single album (not individual tracks)
• DO NOT PROCEED until user confirms

Phase 2: NAS Staging

Step 9: Archive MP3s → Invoke: Skill("media-management:archive-media", args="mp3 $ARTIST $ALBUM")

• Source: $LIBRARY_STORAGE/Artist/Album/ (from Apple Music, captures edits)
• Destination: $ARCHIVE_WORKDIR/to_nas/mp3/Artist/Album/

Step 10: Verify MP3 archival

• Check the JSON output from archive-media: verified should be true
• If not, warn user about count mismatch

Phase 3: WAV Processing

Step 11: Extract WAV ZIP

Run the extraction script:

bash scripts/extract-zip.sh "$WAV_ZIP" "$DOWNLOADS/$RELEASE_NAME-wav"

Step 12: Archive WAVs (SKIP Apple Music) → Invoke: Skill("media-management:archive-media", args="wav $ARTIST $ALBUM $WAV_EXTRACTION_FOLDER")

• WAVs NEVER go through Apple Music import

Step 13: Cleanup → Invoke: Skill("media-management:cleanup", args="$ARTIST - $ALBUM")

• Moves original ZIPs to processed/, cleans extraction folders