lh-focus v1.1A

Immutable. This exact content never changes and is served at /api/v1/blob/b3e48a8c5330390e.

---
name: lh-focus
description: >
  The discipline of saying no: killing features, cutting product lines, finding
  the one thing, and shipping with practical sense. Trigger: load when the
  question is what to keep and what to cut — a roadmap, backlog or feature list
  (an opinion on one included), everything marked top priority, a product
  spreading into adjacent products, or scope colliding with a date.
license: Apache-2.0
metadata:
  author: jrodrigopuca
  version: "1.1"
  local-agents-id: "sk-99599e"
---

## Critical Patterns

### 1. Focus is subtraction with a spine

Focus is not picking favorites — it's saying no to ideas that are GOOD, that
someone loves, that would probably work, because they are not THE thing. The
1997 exercise is the template: Apple's product line went from dozens to a 2x2
grid (consumer/pro × desktop/portable), and the company survived because of
what it STOPPED doing. When a portfolio/roadmap arrives, draw its version of
the grid first: what are the 2-4 boxes, and what falls outside every box?

### 2. The top-10-then-cut-7 ritual

For any planning conversation: list the ten things the product could do next,
ranked. Then cross out the bottom seven — they are now explicitly NOT being
done, said out loud, with names. The unranked backlog is where focus goes to
die, because everything in it is still secretly alive. Three priorities that
everyone can recite beat ten that live in a spreadsheet.

### 3. Every feature pays rent in the one sentence

The product has one sentence (from
[lh-brutal-critique #5](../lh-brutal-critique/SKILL.md)). Every feature must make
that sentence MORE true or it's a candidate to kill — including features that
are polished, shipped, and someone's baby. Audit questions per feature: does
removing it make the product harder to love, or just longer to demo? Would a
new user miss it in week one? Does it exist because users love it or because
removing it requires a meeting?

### 4. Cut scope, never quality — and rarely the date

When the date and the plan collide, the order of sacrifice is fixed:

1. **Scope** — ship fewer things, each one finished and great
2. **Date** — only when the missing piece IS the product (a music player
   without playback isn't "an MVP", it's nothing)
3. **Quality** — never on what ships (judgment #7; this list is its order of
   sacrifice). Quality is the user's experience of the scope that remains: it
   works, it handles its errors, it survives the second week. A refinement
   nobody will feel this release is scope, and cutting it is step 1.

"Finished" means the whole experience — empty states, errors, the second week
— not the demo path. Half of a great product is a great half-product; all of a
mediocre product is just mediocre.

### 5. Say no with the reason attached

A "no" without a why breeds resentment and re-litigation. The formula: "No,
because we are betting everything on X, and this dilutes that bet — it goes to
the parking lot, and here's what would change my mind." Keep the parking lot
real (revisit it at each planning cycle) so "no" means "not now, and honestly,"
not "I didn't listen".

### 6. Practical sense: vision with a ship date

The visionary's paradox, resolved: dream in years, ship in months. Every big
vision gets decomposed into shippable products — each one whole, lovable, and
useful on its own, each one a step toward the vision (the iPod before the
iPhone before the iPad; the vision was decades, the ships were quarters). A
vision with no next ship is a poster; a ship with no vision is a treadmill.
Hold both or hold neither. You name the first ship's SHAPE — whole, lovable,
useful alone; which one goes first, and when, is
[product-manager](../../../product-manager/AGENTS.md)'s discipline, and a
visionary who writes the dates has stopped being the one who says no.

## Resources

- Sibling skills: [lh-brutal-critique](../lh-brutal-critique/SKILL.md) (the verdict
  that feeds the kill list), [lh-inspire](../lh-inspire/SKILL.md) (selling the "no"
  as the path to the great)
- Decomposition mechanics: [generalist/lh-decomposition](../../../generalist/skills/lh-decomposition/SKILL.md)
  (vertical slices are the engineering face of "ship whole products")