agentleFS
Sign inStart free
hamzafarooqhamzafarooq/claude-code-starter/.claude/skills/sharpen/SKILL.md

sharpen

Skill145 starsChanged 6 months agoMITView on GitHub

Give a workplace message an expert second read before you send it. Paste a draft email, Slack message, or exec update — get a signal-strength rating, a critique of exactly what's weak and why, and a ready-to-send rewrite. Use when the user wants feedback on a draft message, or asks to make something clearer, sharper, or more likely to get a yes.

We split this skill into the rules it gives an agent

7 rules. 0 are common, 7 are distinctive, and 0 are very specific to this project.

  • 0

    Common

    Most of the 15 edited versions we read keep these.

  • 7

    Distinctive

    This author's own choices. Worth a look.

  • 0

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    If the draft came in with the command, don't ask for it again.

    line 10
  2. Distinctive:

    Before writing anything, run the draft through the six checks.

    line 12
  3. Distinctive:

    Never add facts, numbers, names, or commitments that weren't in the draft.

    line 66
  4. Distinctive:

    Keep the user's voice. Sharper, not different.

    line 67
  5. Distinctive:

    Don't over-format. A Slack message is not a memo — no headers, no bullets unless the original had them.

    line 68
  6. Distinctive:

    If the draft is already strong, say so and make small edits only.

    line 69
  7. Distinctive:

    Ask: "Want a shorter version, or one aimed at a different reader — your manager vs.

    line 72

“Common” is measured: kept by at least 6 in 10 of the 15 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Ask the user:

  1. "Paste the draft — email, Slack message, exec update, anything you're about to send."
  2. "Who's reading it, and what do you want them to do after reading?"

If the draft came in with the command, don't ask for it again. Only ask question 2 if the reader and the goal aren't obvious from the draft itself.

Before writing anything, run the draft through the six checks. Answer each one honestly.

The six checks:

  1. Lead — Is the most important sentence first? Or is it buried under warm-up, apology, and background?
  2. Ask — Is there one clear request the reader can say yes or no to in a single reply? Is it near the top?
  3. Context — Can the reader act without sending a follow-up question? And is there background they already know that's just taking up space?
  4. Hedges — Count them: "just", "I think", "maybe", "might", "sort of", "if possible", "no worries if not", "I could be wrong but", "quick question". Each one tells the reader you're not sure you should be asking.
  5. Objections — What's the first thing the reader will push back on? Does the draft answer it, or leave it hanging?
  6. Length — Could this be half as long and lose nothing? Every sentence has to earn its place.

Then produce:

Sharpen

Signal strength: 🔴 Weak / 🟡 Mixed / 🟢 Strong

One sentence on why. E.g. "The ask is clear, but it's the last line of a 200-word email and most readers won't get there."

What's weak, and why

Go check by check. Only list the checks that flagged something. For each one, quote the exact words from the draft, then say what they cost the reader.

  • Buried ask — "[quoted line]" → This is the whole point of the message and it's in paragraph 4. The reader has to dig to find out what you want.
  • Hedge words — "[just wanted to]", "[I think maybe]" → Three hedges in two sentences. It reads like you expect a no.
  • Missing context — The reader doesn't know [X], so they can't answer without asking you first.
  • Unaddressed objection — The first thing [reader] will think is "[objection]". The draft doesn't answer it.
  • Weak lead — "[opening line]" → Warm-up. The real message starts three sentences later.
  • Too long — [N] words to make one point. The reader skims and misses the ask.

What's working

One or two things to keep. Be specific — "the second paragraph gives the exact number they need" beats "good tone."

Rewrite

[The ready-to-send version. Same voice, same facts, no invented details.]

Rules for the rewrite:

  • First sentence is the point or the ask
  • The ask is explicit, with a deadline if one exists
  • Zero hedge words
  • The likely objection is answered in one line
  • Roughly half the length of the original, unless the original was already tight
  • Match the channel — Slack gets no greeting or sign-off, email gets a one-line opener

What changed

3 bullets max: what moved, what got cut, what got added. The goal is that the user learns the pattern, not just gets this one fix.

Rules:

  • Never add facts, numbers, names, or commitments that weren't in the draft. If context is missing, flag it in the critique and put a [placeholder] in the rewrite.
  • Keep the user's voice. Sharper, not different. If they write casually, the rewrite is casual.
  • Don't over-format. A Slack message is not a memo — no headers, no bullets unless the original had them.
  • If the draft is already strong, say so and make small edits only. Don't manufacture problems to seem useful.
  • Rate honestly. Most first drafts are 🟡. A 🟢 should be earned, not handed out.

Ask: "Want a shorter version, or one aimed at a different reader — your manager vs. the whole team?"

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

Check your setup, free