# Headroom > Fix work that leaves no room for the person who has to use it, built on the logic of making a beat someone else has to rap on - the mute test, collision mapping, and carving space instead of turning things up. Use whenever the user is building something another person performs on top of - a sales deck a rep presents, a template someone fills in, brand guidelines a team works under, slides for a speaker, onboarding, a workshop, a partner pitch, a form, a curriculum. Also use when a deck upstages its presenter, a template fights the person using it, guidelines feel restrictive, reps go off script, a page is busy, or a thing is good on its own and awkward in use. - Skill: `michaeldoesmarketing/headroom` (Agent Skill, multi-file: 3 files) - Install (CLI): `npx skillmds@latest add michaeldoesmarketing/headroom` - Raw SKILL.md: https://api.skillmd.com/api/skills/michaeldoesmarketing/headroom/raw - Safety review: pending - Works with: Claude Code, Claude.ai, OpenAI Codex - Category: Docs & Writing - Author: MichaelDoesMarketing (https://skillmd.com/u/michaeldoesmarketing) - Updated: 2026-09-17 - Page: https://skillmd.com/skills/michaeldoesmarketing/headroom --- # Headroom The beat that sounds best on its own is usually the worst beat to rap on. Every frequency is occupied, every bar has an event, and there is nowhere for a voice to sit. Producers who only make beats that hit in the car make records nobody can perform on. An enormous amount of work is exactly this and gets built the same wrong way. Sales decks that leave the rep nothing to say. Templates that fight whoever fills them in. Brand guidelines that strangle the teams working under them. Slides that upstage the speaker. All of it built to be impressive alone, by someone who forgot a human has to perform on top of it. The fix is not making things smaller. It is deciding, element by element, who occupies which channel. Use the logic, do not perform the frame. Never talk about beats or mixing in the output unless the user asks where this comes from. ## The two mechanisms **1. Move, do not amplify.** When two things compete for the same space, the fix is never to make one louder, bigger, or bolder. It is to move one out of the way. Any recommendation that adds emphasis to resolve a collision is the wrong recommendation. **2. The mute test.** Take an element out. Ask what the performance is like without it. If it is better or the same, that element was taking room rather than giving it. Run it on everything, and run it hardest on the thing the user is proudest of. Read `references/mute-test.md`. ## Step 1: Name the performer and the performance Most artifacts that fail this were built without a specific person in mind. Establish: - **Who uses this.** A named role at minimum, a named person if possible - **In what situation.** Live in a room, over video, alone at a desk, under time pressure, in front of a hostile audience - **Doing what.** Presenting, filling in, adapting, teaching, deciding, selling - **How good they are at it.** This changes every answer that follows If there is no performer, say so and stop. Some things are consumed directly with no human on top: a self-serve page, an email to a customer, a document read alone. Those need a different judgment and forcing this frame onto them makes them worse. Say that plainly rather than proceeding. ## Step 2: Set the budget Every performance has a fixed total: time, attention, cognitive load, physical space. Work out roughly how much exists, and how much the performer needs for their part. An artifact using ninety percent of a budget has left ten, and anything the human adds pushes past the limit and distorts rather than improves. This is the actual meaning of headroom. Space deliberately left before the ceiling. ## Step 3: Map collisions Six channels a performer needs. For every element in the artifact, mark which ones it occupies, then check which of those the performer also needs. Overlap is a collision. - **Words** - what gets said. If the slide contains the sentence they were going to say, you took it - **Timing** - how long they have. A dense artifact eats their clock before they open their mouth - **Decisions** - what they get to choose. Over-specification removes judgment and the person becomes a reader - **Attention** - the audience looks at the artifact or at the human, never both - **Credit** - who gets to look smart. An artifact that does all the thinking turns the performer into a narrator of someone else's work - **Adaptation** - room to respond to what is actually happening in the room `references/channels.md` covers each one, what it looks like when it collides, and who should usually win. ## Step 4: Run the mute test Element by element. Remove it, describe the performance without it, and mark one of: - **Worse.** The element is load-bearing. It stays - **Same.** It was taking room for nothing. Cut it - **Better.** It was actively in the way. Cut it and note what improved Then run it on the element the user likes most, and report the result honestly whatever it is. ## Step 5: Carve For each collision, decide who keeps the channel, then move the loser out rather than turning the winner up. Standard carves, covered in `references/collisions.md`: - Sentence on the slide the speaker will say out loud, so cut the sentence to a fragment or an image - Instruction in a template duplicating what the field name already says, so cut the instruction - Guideline specifying an outcome when it should specify a boundary, so restate it as a limit rather than a prescription - Two elements making the same point, so merge or drop one **Never fill the room you cleared.** A pass that removes three things and adds two has done nothing. The space is the deliverable. ## Step 6: Leave the hooks Clearing space is half the job. The other half is designing where the person goes. A good beat has pockets built for the verse, not just absence. An artifact with room should have deliberate openings: the place the rep tells their own story, the field where the user's judgment goes, the slide that exists to be talked over, the section a local team is expected to replace. Mark them, and make it obvious they are openings rather than omissions. An unmarked gap reads as an oversight and gets filled by someone trying to be helpful. Read `assets/hooks.md`. ## Modes **Clear the mix** - the default. The artifact exists, find the collisions, carve, mark the hooks. **The mute test** - the user is attached to something and wants to know if it earns its place. Run step 4 alone, starting with their favorite element. **Build for a voice** - making it from scratch for a known performer. Steps 1, 2, 6, then construct. **House band** - the artifact will be used by many different people whose skill varies, and the user does not know who. Templates, guidelines, enablement material. Different problem entirely, covered in `references/performers.md`. ## Output format Deliver in this shape unless the user asks for something else. Prose, no long preamble. ``` THE PERFORMER Who plays on this, in what situation, and how much room they need. THE COLLISIONS Each one: the element, the channel it takes, and who should have it. THE MUTE TEST What got muted and what happened. Include their favorite. THE CARVE What moves, what goes. Move, never amplify. THE HOOKS Where the person is supposed to go, marked so it reads as design. WHAT STAYS Elements that survived, and why. Not everything is in the way. ``` The worksheet is `assets/mix-sheet.md`. ## Rules of engagement **Move, never amplify.** Any fix that resolves a collision by adding emphasis is the wrong fix. **Never fill the space you cleared.** Removing three and adding two is a lateral move dressed as work. **Not everything has a performer.** When the artifact is consumed directly with no human on top, say so and decline the frame rather than forcing it. **More room is not always better.** A nervous first-time presenter needs scaffolding a veteran would find suffocating. Over-clearing for someone who cannot fill the space is its own failure, and it is the failure this skill will cause if it is run without step 1. **Some things must be said by the artifact.** Numbers, legal language, anything that has to survive being misremembered or repeated wrong. Those keep their channel regardless of collision. **The performer's ego is not a channel.** The goal is the performance working, not flattering whoever is on stage. Sometimes the artifact should carry more, and a speaker who wants everything stripped so they can freestyle over nothing is asking for a worse outcome. **Say what stays.** An output that only lists what is in the way gives a false picture of the work and will not get used twice.