skill CX Codex

sharpen — Claude Code skill: deepen a task and rewrite the prompt

Q: How do I install sharpen in Claude?

Run `npx skills add backnotprop/pstack --skill sharpen -a claude-code`. The skill file lands at `~/.claude/skills/sharpen/SKILL.md`.

Q: Does the skill run on its own?

No. It ships with `disable-model-invocation: true`, so the model never calls it by itself — only when you type `/sharpen`.

The skill runs your task through four questions and rewrites the prompt so the work goes deeper instead of wider.

Install into Claude Code · Skills CLI (Vercel) All methods and platforms ↓

npx skills add backnotprop/pstack --skill sharpen -a claude-code

Open on GitHub

What sharpen does

You hand the agent a task, and it looks at that task the way a professional in the field would. First it says what a specialist would do here that you never asked for. Then which raw material they would start from instead of the already-digested version. Next come the specific methods and tools, and last — what besides you two would show the result is right.

After that the skill rewrites your prompt to ask for exactly those things. Choices that belong to you stay open; it does not make them for you. Inside there is a single SKILL.md file with the instruction — no scripts, no external dependencies.

Model invocation is off: disable-model-invocation: true means the agent will not trigger the skill on its own, only when you call it. Worth knowing if you expect it to pick itself up mid-conversation.

What it helps with

turn a vague task into a specific one
see what a professional would add to the task
find raw source material instead of summaries
pick methods and tools that fit the job
decide how to verify the result

Example prompts for the agent

/sharpen write a quarterly sales report
/sharpen add tests for the auth module
/sharpen fix the broken import in the parser

How to install sharpen

1Skills CLI (Vercel)

npx skills add backnotprop/pstack --skill sharpen -a claude-code

2Manually: put the files in place

~/.claude/skills/sharpen/SKILL.md

Copy the whole skill folder; for one project use .claude/skills/sharpen/

1Skills CLI (Vercel)

npx skills add backnotprop/pstack --skill sharpen -a codex

2Manually: put the files in place

.agents/skills/sharpen/SKILL.md

For all projects use ~/.agents/skills/

1Skills CLI (Vercel)

npx skills add backnotprop/pstack --skill sharpen -a cursor

2Manually: put the files in place

.cursor/skills/sharpen/SKILL.md

Cursor also reads .agents/skills/ and .claude/skills/

1GitHub CLI

gh skill install backnotprop/pstack sharpen

2Manually: put the files in place

.github/skills/sharpen/SKILL.md

1Skills CLI (Vercel)

npx skills add backnotprop/pstack --skill sharpen -a gemini-cli

2Platform CLI

gemini skills install https://github.com/backnotprop/pstack --path skills/sharpen

Manually: put the files in place

.devin/skills/sharpen/SKILL.md

The legacy .windsurf/skills/ path still works

Skills CLI (Vercel)

npx skills add backnotprop/pstack --skill sharpen -a opencode

Step by step, with commands for each platform: How to install skills in Claude Code: five ways

Questions about sharpen

How do I install sharpen in Claude?

Run npx skills add backnotprop/pstack --skill sharpen -a claude-code. The skill file lands at ~/.claude/skills/sharpen/SKILL.md.

What does sharpen do?

It asks four questions about your task — what a pro would do, what raw material to start from, which methods and tools to use, and how to verify the result — then rewrites the prompt around the answers.

Does the skill run on its own?

No. It ships with disable-model-invocation: true, so the model never calls it by itself — only when you type /sharpen.

Source text

SKILL.md · MIT

---
name: sharpen
description: "Make a task deeper, not bigger, then rewrite the prompt. Use for /sharpen."
disable-model-invocation: true
---

I want a deeper version of this task, not a bigger one. Before rewriting it, do four things:

1. Say what a top professional in this field would do here that I'm not asking for.
2. Name the rawest material they would start from instead of what has already been digested for me.
3. Name the specific methods and tools they would use.
4. Say what, other than you or me, would show the result is right.

Then rewrite my prompt to ask for exactly that.

Leave choices that are mine to make open instead of deciding them.

create-agent-skills glittercowboy A mentoring skill for writing and refining agent skills: SKILL.md structure, router pattern, progressive disclosure. CX Codex +1 503installs bmad-prfaq bmad-code-org This skill tests a product concept with a Working Backwards PRFAQ, moving from a finished-product press release to hard customer and stakeholder questions. CX Codex +1 11,300installs define-goal official openai Shapes vague requests into concrete, measurable goals with explicit pass criteria before an agent starts working. CX Codex +1 2,800installs instinct-system codewithmukesh Captures code patterns, user corrections, and project findings in separate stores. Use it when an agent needs to retain lessons from recurring work. CX Codex +1 2,800installs aim LilMGenius A Claude skill for the moment files, a repo or a chat scrollback arrive with a thin ask. It reads everything and proposes what you meant, so you only confirm. CX Codex +1 1,200installs readchk LilMGenius This skill checks whether the agent understood your instruction before it spends real work on it. Useful for long or hard-to-undo requests. CX Codex +1 1,200installs

From the same set44

principle-foundational-thinking A principle set for the stage before logic: pick core types and data structures, order the scaffold, check what concurrent actors share. CX Codex +1 5,100installs correct Finds repeated mistakes made by coding agents and makes them impossible through architecture, types, linters, or tests. Useful when agents keep making the same wrong edits. CX Codex +1 4,200installs benchmark-checklist A skill that makes the agent answer seven questions with evidence from a run before it reports a speedup or a regression. CX Codex +1 4,100installs principle-explain-the-number A skill for the moment before you trust a measured number: name what caps it, then rule out that the test measured something other than your work. CX Codex +1 4,100installs poteto-help Answers questions about pstack: how to set it up, how to run /poteto-mode, which skill fits a task. It hands you a prompt instead of starting work. CX Codex +1 3,700installs ketchup Summarizes chat updates since your last message in plain words and visual diagrams, highlighting decisions that need your approval. CX Codex +1 2,900installs create-formal-verification-skill This skill selects code for formal verification and creates a project-local guide for checking it again after changes. Use it when tests leave important invariants or event orderings uncovered. CX Codex +1 1,200installs maintain-formal-verification-skill Checks whether a project’s formal verification still matches changed code, then reruns it and classifies failures. Useful after merges or refactors that bypassed the usual verification workflow. CX Codex +1 1,100installs