La interfaz está localizada; las descripciones y documentos de las Skills se traducen automáticamente con IA a su idioma. El código, los comandos, rutas y términos técnicos se mantienen en el original, y prevalece el documento del autor.
Explorar Skills
Productividad · SKILL.md

mastering-prep

mastering-prep

Use when the user is finishing a track and wants to check it's ready to send to a mastering engineer or for self-mastering. Audits headroom, peak levels, mono compatibility, frequency balance, LUFS. Examples - "is this ready to master?", "audit my mix before export", "check my levels".

Independent GitHub snapshot · Automated classification · Original documentation governs

Documentación de la Skill

GitHub

name: mastering-prep description: Use when the user is finishing a track and wants to check it's ready to send to a mastering engineer or for self-mastering. Audits headroom, peak levels, mono compatibility, frequency balance, LUFS. Examples - "is this ready to master?", "audit my mix before export", "check my levels".

Mastering Prep

Pre-master audit. Confirm the mix is in a clean state for either external mastering or self-mastering. Does NOT apply final mastering — that's a separate workflow with explicit user intent.

Workflow

1. Read the master bus

  • get_track_info for master
  • get_device_parameters for any plugins on master
  • Check current peak level, RMS, LUFS if available

2. Run the audit checklist

For each item, mark ✅ pass or ❌ fail with the specific reading:

CheckTargetWhat to look for
Peak level-3 to -6 dBMaster should have headroom — never above -1 dB
No master limiter active for export-to-masteringOffMastering engineer wants raw mix; remove limiter from master
No master compressor over 2 dB GR<2 dBHeavy mastering compression on a pre-master is destructive
LUFS integrated-16 to -20 LUFS for pre-masterToo loud = no headroom for mastering
Mono compatibilityAll elements audible in monoToggle Utility's mono switch — verify nothing disappears
Phase coherenceAbove 0.0 (positive)Use Spectrum or Stereo & Phase — dips below 0 = destructive cancellation
Frequency balanceNo spikes >6 dB above neighborsBass vs mids vs highs evenly spread
Sub-bass below 40 HzCut or controlledAnything below 30 Hz wastes headroom and clouds masters
Above 18 kHzCut if no musical contentTames hiss/artifacts; saves headroom
DC offsetNoneUse Utility's DC filter on the master
Track-level clippingNoneInspect each track's meter — yellow is fine, red is not
Inactive devicesBypass or removeUnused devices on tracks waste CPU and confuse archiving

3. Identify and report

Print the audit table with results. Lead with failures, then warnings, then passes.

"Audit results: ❌ Master peaking at -0.8 dB — needs at least 3 dB more headroom ❌ Vocals on track 3 mono-incompatible — disappears 30% in mono fold ⚠️ LUFS integrated -10.2 — too loud for pre-master, drop master fader 4 dB ✅ Phase coherence 0.7 — good ✅ No DC offset ✅ Sub-bass clean below 30 Hz"*

4. Propose fixes

For each failure, propose a specific fix. Apply only on user confirmation:

  • Master peaking → drop master fader by N dB
  • Mono incompatibility → identify the offending track (usually a stereo widener or M/S processing pushed too far) and reduce
  • LUFS too high → master fader trim
  • DC offset → add Utility DC filter

5. Final export checklist

If user is exporting now, also confirm:

  • ✅ Render quality: 24-bit, 44.1 or 48 kHz (matching project sample rate)
  • ✅ Render mode: Real-time off (faster, identical output for non-resource-bound projects)
  • ✅ Dither: ON for 16-bit exports, OFF for 24-bit (24-bit goes to mastering raw)
  • ✅ Normalize: OFF (mastering will handle this)
  • ✅ Convert to Mono: OFF (unless single-mono output requested)
  • ✅ File name: includes track name, BPM, key, version (e.g., track-name-95bpm-Cm-v3.wav)

Don'ts

  • Don't apply mastering chain on the user's behalf. This skill audits, doesn't master.
  • Don't recommend external mastering tools without asking budget and intent.
  • Don't normalize. Normalize = peak-based loudness, irrelevant to perceived loudness, breaks headroom.
  • Don't auto-fix more than one issue at a time — let the user listen between each.
Skills guardadas