More screenshots
Works with the AI Agents you already use
Mix Engineer: Stem Cleanup and Remix
Clean and balance generated or imported audio before mastering: stem triage, defect repair in a fixed order, EQ cleanup, vocal-first balance, and a headroom-safe handoff with a written note.
$5
Mix Engineer: Stem Cleanup and Remix
Example session with this skill installed
My generated indie folk track came back as stems: lead vocal, acoustic guitar, bass, drums, and strings. The vocal is harsh and gets lost in the chorus, the guitar sounds muddy, and there is a click in the intro. What do I do, and in what order?
- Read your context and instructions
- Compiled the mix engineer
Mode: stems. Triage first, nothing processed yet. Listen at three volumes; measure peaks, DC offset, stereo correlation, noise floor. Classify what you reported: the click is a defect, fix unconditionally. Vocal harshness and guitar mud are defect-class cleanup. The vocal getting lost in the chorus is a balance call, fixed last and documented, because it is probably the strings masking the vocal, not the vocal being quiet.
Repair order, in sequence. 1) Hygiene: repair the intro click by interpolation around the damaged samples; trim dead air; fix any DC offset. 2) Low-end cleanup, high-pass per stem: vocal 80 Hz, guitar 90 Hz, bass 25 Hz, drums (kick) 25 Hz, strings 50 Hz, gentle slopes. 3) Mud: the guitar's boxiness lives at 150-400 Hz; sweep a narrow boost to find the worst frequency (expect around 250 Hz) and cut −2 to −3 dB there. One cut only. 4) Harshness: the vocal's edge is in the 2-5 kHz region; tame with a −1 to −2 dB high shelf, not a notch. Check sibilance at 5-9 kHz and de-ess if needed, before compression (compression worsens sibilance). 5) No noise reduction: this audio is generated, it has no room tone, and gating it would chew up breaths and decay tails. 6) Gentle dynamics: lead vocal −15 dB threshold, 2.5:1; bass 3:1 for a consistent bottom; strings slow attack at 1.5-2:1 so they stay a pad, not a competitor.
Balance pass, vocal first at playback volume. Then drums, bass, then guitar and strings around them. Strings: cut −2 dB around 2.5 kHz so the vocal owns the 1-3 kHz presence slot in the chorus; that, not a volume raise, is the fix for "lost in the chorus." Pan guitar 40% one side, strings widest, the core centered. Mono check after panning: vocal and kick must stay solid; if the chorus collapses in mono, the strings were separated by stereo instead of level, so rebalance levels.
Handoff state: peak between −6 and −3 dBFS, no limiter anywhere, 24-bit WAV at the session rate, note reading: click repaired, guitar mud cut at 250 Hz, vocal de-essed pre-compression and shelf-tamed, strings dipped at 2.5 kHz for chorus presence (balance call, documented). Rubric target 10+ of 14, no zero in defects, mono, or headroom. Ready for mastering.
Connects securely to your tools. The creator never sees your data.
About this skill
This skill is the mix polish stage between raw audio and mastering. Its rule is fix, don't create: it removes defects, cleans tone, and balances what exists, and it sends arrangement or performance problems back to production instead of patching them for hours. It works in two modes, separated stems (preferred) or a full stereo mix (broader, gentler moves), and never both, because running stem chains and full-mix processing double-applies everything.
Every track starts with triage: listen at three volumes, measure peaks, DC offset, stereo correlation, and noise floor, then classify every problem as a defect (fix unconditionally) or a balance call (fix and document the judgment). Processing follows a fixed repair order: edits and hygiene, low-end cleanup with per-stem high-pass values, mud and boxiness cuts, harshness and sibilance (de-essing before any compression), noise reduction on recorded material only, then gentle per-stem compression with starting-point settings per instrument.
The balance pass builds vocal-first, then drums, bass, and everything else, with a panning map, frequency slot assignments so stems stop fighting, and a mandatory mono check. A readiness rubric scores the finished mix, and the handoff to mastering is explicit: peak between −6 and −3 dBFS, no limiting, 24-bit WAV, and a note covering what was fixed, what was deliberately left, and anything flagged. It also covers the bleed caveat for AI-separated stems, where surgical moves on one stem damage whatever leaked into it.
Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse All Claude Skills.
This skill mixes; it stops before mastering. Loudness targeting and final limiting belong to the mastering stage, and mixes are deliberately delivered un-limited with headroom. It cannot repair a fundamentally broken stem: clipping baked into a lead vocal or a performance that falls apart gets flagged back for regeneration rather than patched. On AI-separated stems, bleed between stems limits how surgical any fix can be, and the skill says where that line is.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before buying a skill.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Money-back guarantee
Doesn't do what its page promises? You get your money back.
Trust & safety
Security scan
Security scan passed 24 days ago
Permissions
No special permissions declared or detected