sl-builds

Sharpen Claude Prompt Refiner — AI skill for Claude Code

AI community

A Claude skill that rewrites vague prompts into tight specs before any work starts.

How to install Sharpen Claude Prompt Refiner

This entry records only its repository, not the path inside it, so there is no exact command to give. Open sl-builds/sharpen-Claude-Prompt-Refiner and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

What Sharpen Claude Prompt Refiner does

A Claude skill that rewrites vague prompts into tight specs before any work starts. Locks the target, kills hallucinated facts, forces the shortest correct path, and bans default AI aesthetics with a pre-ship slop audit. Fewer tokens, fewer wrong answers, output that does not look machine-made.

Alternatives in AI

Surface — Show ranked attack surface for a target from its recon manifest + hunt memory 3.8k ★
Claude Skill Web Clone — Methodology for faithfully cloning any website (static / React / WebGL) — without copying AI-hallucinated code 969 ★
Aurora — Aurora - independent community smart home orchestrator 103 ★

README

sharpen

A Claude skill that rewrites a vague prompt into a tight spec before any work starts, then makes the model take the shortest correct path to it.

Built to fix four things that waste the most tokens:

Wrong target — the work is correct but not what you wanted
Invented facts — confident output with nothing checkable under it
Wasteful path — the right answer reached the long way round
Default AI look — correct, efficient, and instantly recognisable as machine-made

What it does

Before touching the task, Claude reduces your request to five lines and shows them, then starts work in the same reply:

If something is missing, it looks it up rather than asking. It only asks when a wrong guess would waste the whole run, and never more than two questions.

For anything visual it commits to a stated design direction and runs a slop audit before shipping: no purple gradients, no glassmorphism, no emoji bullets, no grids of identical rounded cards, no "unlock" or "seamless".

Install

**Claude Code**

git clone https://github.com//sharpen.git ~/Projects/sharpen
mkdir -p ~/.claude/skills
ln -sfn ~/Projects/sharpen ~/.claude/skills/sharpen

**Claude app** — open Skills settings and add `SKILL.md`.

Use

It triggers on its own for vague, multi-part or visual requests. To force it:

Also fires on "refine my prompt", "be efficient", "don't hallucinate", "no AI slop".

Not for

Small, unambiguous questions. A one-line answer should stay a one-line answer, and adding a spec block to trivial work is the same waste the skill exists to prevent.