Help us improve
Share bugs, ideas, or general feedback.
From chief-skills
Cuts over-engineered plans down to one achievable increment by reading the codebase, identifying core value, and sorting work into keep/defer/drop.
npx claudepluginhub thaitype/chief --plugin chief-skillsHow this skill is triggered — by the user, by Claude, or both
Slash command
/chief-skills:slim-downThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
Take a plan or design that's grown too big and cut it down to **one increment** — the smallest step that delivers value and doesn't block future growth.
Prevents feature creep with checklists for validating user needs, measuring impact, assessing complexity, defining MVPs, and managing scope in software projects.
Generates tightly scoped implementation plans (≤5 steps, ≤1250 words) for tasks, framed as staff engineer discussions. Use for sprint-ready breakdowns.
Share bugs, ideas, or general feedback.
Take a plan or design that's grown too big and cut it down to one increment — the smallest step that delivers value and doesn't block future growth.
Make it work before make it great — but never paint yourself into a corner.
Cut scope, not corners. Remove features, not quality. The increment should be small but extensible — no dead-ends, no "throw it away later."
Before cutting anything, understand what already exists:
Summarize in a short block:
Ask: "What is the one thing this plan must deliver to be useful?"
Draft a single sentence describing the core value. Show it to the user. This is the anchor — everything else gets judged against it.
Take every item from the plan and sort it:
| Criteria | |
|---|---|
| Keep | Required for the core to work. Can't ship without it. |
| Defer | Valuable but not needed yet. Design so it can be added later. |
| Drop | Nice-to-have, speculative, or solves a problem that doesn't exist yet. |
Show the table. Let the user move things around.
Rules for sorting:
For each Defer item, verify:
If yes → move that item back to Keep or adjust the design.
If the user asks "will this scale?" — answer honestly based on the code, not hypotheticals.
Co-write a short definition of what to build:
Write to a file if the user wants. Otherwise just confirm alignment.