resume
정보
이 스킬은 퍼지 이름 매칭을 통해 기존 음악 앨범을 찾아 상세 상태와 다음 단계를 표시함으로써 작업을 재개합니다. 사용자가 앨범 이름을 언급하거나 이전 작업을 이어서 진행하려 할 때 트리거되며, MCP 도구를 활용해 프로젝트를 찾고 캐시 상태를 처리합니다. 앨범을 찾을 수 없거나 여러 개가 매칭될 경우 명확한 복구 경로를 제공합니다.
빠른 설치
Claude Code
추천npx skills add bitwize-music-studio/claude-ai-music-skills -a claude-code/plugin add https://github.com/bitwize-music-studio/claude-ai-music-skillsgit clone https://github.com/bitwize-music-studio/claude-ai-music-skills.git ~/.claude/skills/resumeClaude Code에서 이 명령을 복사하여 붙여넣어 스킬을 설치하세요
문서
Resume Album Work
Purpose: Find an album and resume work where you left off.
Usage:
/bitwize-music:resume <album-name>
/bitwize-music:resume my-album
/bitwize-music:resume "demo album"
When to use: When user wants to continue working on an existing album.
Instructions
When this skill is invoked with an album name:
Step 1: Find the Album via MCP
- Call
find_album(name)— fuzzy match by name, slug, or partial (case-insensitive) - If not found: MCP returns available albums — suggest closest match or
/bitwize-music:new-album - If multiple matches: list all with paths, ask user which one
- If MCP returns stale/missing cache error: call
rebuild_state()then retry
Step 2: Get Album Progress
- Call
get_album_progress(album_slug)— returns track counts by status, completion percentage, and detected workflow phase - Call
list_tracks(album_slug)— returns per-track details (status, has_suno_link, sources_verified)
Step 3: Update Session Context
Call update_session(album=album_slug, phase=detected_phase) to set the active album and phase.
Step 4: Determine Current Phase
Based on album and track statuses, identify the workflow phase:
| Album Status | Track Statuses | Current Phase |
|---|---|---|
| Concept | Most "Not Started" | Planning - Need to fill in album README and create tracks |
| Research Complete | Some "Sources Pending" | Verification - Need human verification of sources (documentary albums) |
| Sources Verified | All sources verified | Ready to Write - Sources cleared, begin lyrics (documentary albums) |
| In Progress | Mixed, some "Not Started" | Writing - Need to complete lyrics (or route instrumental tracks to suno-engineer) |
| In Progress | Some "Sources Pending" | Verification - Need human verification of sources |
| In Progress | All have lyrics | Ready to Generate - Run Ready to Generate checkpoint |
| In Progress | Some "Generated" | Generating - Continue generating on Suno. Check Generation Logs for rejected tracks needing regeneration |
| In Progress | All "Generated", none "Final" | Review & Approve - Listen to generated tracks, mark keepers with ✓, regenerate rejected ones |
| Complete | All "Final" | Mastering - Ready to master audio |
| Released | All "Final" | Released - Album is live |
Note: Non-documentary albums skip Research Complete and Sources Verified — they go directly from Concept → In Progress.
Step 5: Report to User
Present a clear status report:
📁 Album: [Album Title]
Location: {content_root}/artists/{artist}/albums/{genre}/{album}/
Status: [Album Status]
📊 Progress:
- Tracks: [X completed / Y total] ([N vocal, M instrumental])
- Not Started: X
- In Progress: Y
- Generated: Z
- Final: N
📍 Current Phase: [Phase Name]
✅ What's Done:
- [List completed items]
⏭️ Next Steps:
1. [Specific action 1]
2. [Specific action 2]
3. [Specific action 3]
Ready to continue? Tell me what you'd like to work on.
Step 6: Recommend the Single Best Next Action
Pick ONE clear recommendation from the decision tree below. Don't list 5 options — pick the best one, include the skill name, and be specific about which track.
Decision Tree (evaluate top-to-bottom, first match wins):
Instrumental detection: Check each track's frontmatter for instrumental: true or Track Details table for **Instrumental** | Yes. Instrumental tracks skip the lyrics workflow entirely and go straight to /bitwize-music:suno-engineer.
Album Status = "Concept"
→ "Define the album concept. Run /bitwize-music:album-conceptualizer"
Album Status = "Research Complete"
→ Any tracks Sources Pending?
YES → "Sources need verification. Run /bitwize-music:verify-sources [album]"
NO → Any "Not Started" tracks instrumental?
YES → "Create Style Box for instrumental track [name]. Use /bitwize-music:suno-engineer"
NO → "Ready to write! Pick a track and use /bitwize-music:lyric-writer"
Album has tracks with "Not Started"
→ Is the first not-started track instrumental?
YES → "Create Style Box for [track]. Use /bitwize-music:suno-engineer directly (instrumental track)"
NO → "Write lyrics for [first not-started track]. Use /bitwize-music:lyric-writer"
Album has tracks with "In Progress" (lyrics partially written)
→ "Finish lyrics for [first in-progress track]. Use /bitwize-music:lyric-writer"
Album has tracks with "Sources Pending"
→ "Verify sources for [track]. Run /bitwize-music:verify-sources [album]"
All tracks have lyrics (or Style Box for instrumentals), none generated
→ Mixed album (vocal + instrumental)?
YES → "All tracks ready! Run /bitwize-music:pronunciation-specialist on vocal tracks, then /bitwize-music:lyric-reviewer, then /bitwize-music:pre-generation-check to validate all gates (instrumental tracks auto-skip lyrics gates)."
NO → "All lyrics complete! Style prompts should be ready. Run /bitwize-music:pronunciation-specialist to check for pronunciation risks, then /bitwize-music:lyric-reviewer for final QC, then /bitwize-music:pre-generation-check to validate all gates before generating on Suno."
Some tracks generated, some not
→ Any Generated tracks without ✓ in Generation Log Rating?
YES → "Track [name] was generated but not approved. Listen and decide:
- Happy? Mark ✓ in Generation Log and set Status: Final
- Not happy? Log the reason, then:
Style issue → /bitwize-music:suno-engineer to revise Style Box
Lyrics issue → /bitwize-music:lyric-writer to fix, then regenerate
Bad luck → Regenerate on Suno with same settings (it's non-deterministic)"
NO → "Generate [first un-generated track] on Suno. Use /bitwize-music:suno-engineer"
All tracks generated, none Final
→ "All tracks generated! Listen to each track and approve:
- Mark keepers with ✓ in Generation Log Rating column
- Reject and regenerate any that don't meet quality standards
- Once all have ✓, batch-approve:
Use update_track_field(album_slug, track_slug, 'status', 'Final') for each approved track.
Once all Final, album advances to Complete automatically."
All tracks generated, some Final
→ Any Generated (non-Final) tracks without ✓?
YES → "Review track [name] — listen and approve (✓) or regenerate"
NO → "All tracks approved! Batch-approve: update_track_field(album_slug, track_slug, 'status', 'Final') for each.
Then import audio with /bitwize-music:import-audio, then master with /bitwize-music:mastering-engineer"
All tracks Final
→ "All tracks approved! Import audio with /bitwize-music:import-audio, then master with /bitwize-music:mastering-engineer"
Album Status = "Complete"
→ "Album is complete! Release with /bitwize-music:release-director"
Album Status = "Released"
→ "This album is released! Consider /bitwize-music:promo-director for promotional content"
→ Also suggest: "Start a new album? Check /bitwize-music:album-ideas list"
Format the recommendation as:
RECOMMENDED NEXT ACTION:
[Clear, specific instruction with skill name and track name]
WHY:
[One sentence explaining why this is the right next step]
When No Album Specified (No Arguments)
If invoked without an album name:
- Call
get_session()— checklast_albumfrom session context, resume that album - If no session context, call
list_albums()to find all in-progress albums - Prioritize: closest to completion > unblocked work > last worked on
- If no albums exist, suggest
/bitwize-music:new-album
Present a multi-album summary if multiple are in progress:
You have X albums. Here's the most actionable:
PRIORITY 1: [album-name] ([genre])
Status: [status] | Progress: [X/Y tracks]
→ [Recommended action]
Also in progress:
- [album-2] — [brief status]
Examples
Example 1: Album in Writing Phase
/bitwize-music:resume my-album
📁 Album: My Album
Location: ~/bitwize-music/artists/bitwize/albums/rock/my-album/
Status: In Progress
📊 Progress:
- Tracks: 3 completed / 8 total
- Not Started: 3
- In Progress: 2
- Final: 3
📍 Current Phase: Writing Lyrics
✅ What's Done:
- Tracks 1-3 have final lyrics
- Album concept and tracklist defined
⏭️ Next Steps:
1. Complete lyrics for Track 4 (in progress)
2. Complete lyrics for Track 5 (in progress)
3. Write lyrics for Tracks 6-8
Ready to continue? Tell me which track you'd like to work on.
Example 2: Album Ready for Generation
/bitwize-music:resume demo-album
📁 Album: Demo Album
Location: ~/bitwize-music/artists/bitwize/albums/electronic/demo-album/
Status: In Progress
📊 Progress:
- Tracks: 8 / 8 total (all lyrics complete)
- Final: 8
📍 Current Phase: Ready to Generate
✅ What's Done:
- All 8 tracks have complete lyrics
- All lyrics phonetically reviewed
- Suno Style and Lyrics boxes filled
⏭️ Next Steps:
1. Run Ready to Generate checkpoint (I'll verify everything)
2. Start generating on Suno
3. Log generation attempts
Shall I run the Ready to Generate checkpoint now?
Example 3: Album Not Found
/bitwize-music:resume my-album
❌ Album 'my-album' not found.
Available albums:
- demo-album (electronic) - In Progress
- example-tracks (hip-hop) - Complete
Did you mean one of these? Or use /bitwize-music:new-album to create a new album.
Implementation Notes
- Use MCP tools -
find_album,get_album_progress,list_tracks,update_sessioninstead of reading state.json directly - Case-insensitive matching - "Sample-Album" should match "sample-album"
- Handle missing albums gracefully - List what exists, don't error
- Be specific about next steps - Don't just say "continue working", say exactly what to do
- Include full paths - User needs to know where files are located
- Use emojis sparingly - Only for section headers in the report
Model
Use Sonnet 4.5 - This is a coordination/reporting task, not creative work.
GitHub 저장소
연관 스킬
llamaguard
기타LlamaGuard는 폭력 및 혐오 발언 등 6가지 안전 범주에서 LLM 입력과 출력을 조정하기 위한 Meta의 70-80억 파라미터 모델입니다. 94-95% 정확도를 제공하며 vLLM, Hugging Face 또는 Amazon SageMaker를 사용해 배포할 수 있습니다. 이 기술을 사용하여 AI 애플리케이션에 콘텐츠 필터링 및 안전 가드레일을 손쉽게 통합하세요.
cost-optimization
기타이 Claude Skill은 리소스 적정화, 태깅 전략, 지출 분석을 통해 개발자들이 클라우드 비용을 최적화할 수 있도록 지원합니다. AWS, Azure, GCP에서 클라우드 비용을 절감하고 비용 거버넌스를 구현하기 위한 프레임워크를 제공합니다. 인프라 비용을 분석하거나, 리소스를 적정화하거나, 예산 제약을 충족해야 할 때 사용하세요.
quantizing-models-bitsandbytes
기타이 스킬은 bitsandbytes를 사용하여 LLM을 8비트 또는 4비트 정밀도로 양자화하며, 최소한의 정확도 손실로 50-75%의 메모리 감소를 달성합니다. 제한된 GPU 메모리에서 더 큰 모델을 실행하거나 추론을 가속화하는 데 이상적이며, INT8, NF4, FP4와 같은 형식을 지원합니다. 이 스킬은 HuggingFace Transformers와 통합되어 QLoRA 학습 및 8비트 옵티마이저를 가능하게 합니다.
dispatching-parallel-agents
기타이 Claude Skill은 3개 이상의 독립적인 문제를 동시에 조사하고 해결하기 위해 다중 에이전트를 배치합니다. 공유 상태나 의존성 없이 해결 가능한 무관련 장애 시나리오에 맞게 설계되었습니다. 핵심 기능은 병렬 문제 해결로, 각 독립 문제 영역마다 하나의 에이전트를 할당하여 효율성을 극대화합니다.
