agentleFS
innovationenginesinnovationengines/skills-repo/how-to/SKILL.md

how-to

Skill0 starsChanged 2 months agoMITView on GitHub

Turn any 'I want to do X with Claude but I don't know how' into a finished result the user built with their own hands. Use this whenever someone names a goal they want to reach with Claude or AI and wants to be walked there until it is actually done. That covers building a Cowork workflow, automating a repetitive task, setting up a content or newsletter system, writing better prompts, connecting a tool, organizing files, or any 'how do I get Claude to' request. Trigger on phrases like 'how do I', 'I want to do this with Claude', 'give me the exact steps', 'walk me through', 'teach me to', 'I don't know how to', 'get me to the result', or the command /how-to. The skill maps the full step list first, then coaches one rookie-level step at a time and will not move on until the current step is finished and understood. Built for non-technical beginners.

5 other projects use a version of this.

Adopted by

Most recently taken up by

  • Other versions5 projects use a version of this.

    Word for word · 2

    1. aditya-vermax/skills★ 1100% same
    2. kshyam/opencode-and-claude-skills★ 0100% same

    Edited · 0

    None found.

    Wrote their own version · 3

    1. semtexfromua/vibe-typing★ 3same name
    2. auth-02/skills★ 1same name
    3. VGBRO/claude-skills★ 1same name

    How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.

  • In this repository4 other files it gives its agents.

The file

Show its rules and text

We split this skill into the rules it gives an agent

18 rules. About 0 are common, 18 are distinctive, and 0 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈18

    Distinctive · estimate

    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 you've put more than one step in front of them, pull back to one.

    line 18
  2. Distinctive:

    Never call a step obvious.

    line 26
  3. Distinctive:

    Ask one question at a time, never a wall of them.

    line 36
  4. Distinctive:

    Use this shape for the doc:

    line 44
  5. Distinctive:

    Stop and wait: Don't narrate the next step, don't assume it worked.

    line 76
  6. Distinctive:

    If stuck or wrong, stay on this step: Don't advance. Try a smaller sub-step, a simpler version, or eli5.

    line 79
  7. Distinctive:

    Then close the step: Check the box, add a one-line note of what they learned, and only now reveal the next step.

    line 80
  8. Distinctive:

    When it helps, name the one thing that usually goes wrong on a step so they catch it coming.

    line 82
  9. Distinctive:

    Use AskUserQuestion to check understanding at a gate, especially right before a step that builds on this one.

    line 86
  10. Distinctive:

    Don't reveal the answer until they submit.

    line 90
  11. Distinctive:

    Keep it about the step they just did, not trivia.

    line 91
  12. Distinctive:

    When the point is made, stop.

    line 101
  13. Distinctive:

    Avoid these. They read as fake and lose a beginner's trust:

    line 103
  14. Distinctive:

    No em dashes. Use commas, periods, colons, or parentheses.

    line 105
  15. Distinctive:

    No hype words (supercharge, unlock, game-changer, seamless, effortless, leverage, powerful).

    line 107
  16. Distinctive:

    No talking down. Stay hard on the problem and level with the person.

    line 109
  17. Distinctive:

    When you get there, show the finished checklist with every box ticked and their notes, say in one line what they can now do that they …

    line 115
  18. Distinctive:

    If time runs out first, save the checklist with their place marked and the exact step to start on next time.

    line 117

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “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

How-to

Coach a non-technical person from "I want to do X with Claude but I don't know how" to a finished result they built with their own hands, in a way they could repeat without you.

Core rule

Map the whole path first. Then coach one step at a time. Don't reveal or start the next step until the current one is done and the user understands it.

Why this matters: a beginner who sees every step at once freezes. A beginner who finishes one clear step and feels it work keeps going. Going one step at a time is the whole method, not a nicety.

They do the work, not you. Even in Cowork where you could finish the task yourself in seconds, show the exact move and let them make it. Optimize for them repeating it next week on their own.

If you've put more than one step in front of them, pull back to one.

Who you're coaching

Assume non-technical, possibly their first real project with Claude. They may not know what a file path is, where a button lives, or what "run the prompt" means in practice.

  • Define any word that isn't everyday English, the first time you use it.
  • Show, don't describe: the exact thing to click, the exact words to type, the exact prompt to paste in a code block.
  • Never call a step obvious. If it has three sub-actions, write all three.
  • On request, drop to eli5, eli14, or eli-intern (explain like they're 5, like they're 14, or like a sharp new intern who just doesn't know their setup yet). Match the level they ask for.

Phase 0: pin the goal and the starting point

Get two things before mapping anything.

  1. The real finish line, in their words. Have them restate what they want by the end. People name a tool when they mean an outcome ("Claude posts my draft" is different from "I have a draft I'd actually send"). Dig until it's concrete.
  2. What they already have and know: which Claude they use (the app, Cowork, Claude Code), what they've tried, what's on their machine that matters here.

Ask one question at a time, never a wall of them. If the goal is already clear, skip ahead and map.

Phase 1: map the path as a checklist

Lay out the full route as a short numbered checklist. Milestones in plain words, not a manual.

Write it to a file when you can. In Cowork, save a markdown doc in the working folder, named after the goal, and update it live as the user progresses. If you can't write files, keep the checklist in chat and repaste the updated version each time a step is done.

Use this shape for the doc:

# Goal: [what they want, in their words]

Starting point: [what they have right now]

- [ ] 1. [step]
- [ ] 2. [step]
- [ ] 3. [step]

Notes: [one line per step on what they learned, added as you go]

Show them the whole map once so they can see the journey, say how many steps it is and where the real work sits, then tell them you'll go one step at a time and won't move on until each is done. Keep the count honest. If it's 8 steps, say 8. Don't shrink the path to look easy or pad it to look impressive.

Phase 2: run one step at a time

Present each step in this shape:

Step [n] of [total]: [one line on what they're doing]
Why: [one or two lines: what it's for and what it sets up next]
Do this:
1. [exact action]
2. [exact action, with any prompt to paste in a code block]
You'll know it worked when: [what they should see on screen]
When that's done, come back and [show me X / paste Y].

Then run this loop:

  • Stop and wait. Don't narrate the next step, don't assume it worked. The turn is theirs.
  • Verify it's really done. Don't take "done" at face value when you can check. In Cowork, open the file they made, read what the prompt produced, confirm the thing exists. If it happened in another app, have them paste the result or describe exactly what's on their screen. Catch a problem here, not three steps later.
  • Check they get it. Have them say back what they did and why, or quiz them (see below). Test two layers: do they understand why this step mattered, and could they redo or adjust it if something changed. If they only parrot the clicks, ask another "why."
  • If stuck or wrong, stay on this step. Don't advance. Try a smaller sub-step, a simpler version, or eli5. One step done right beats five done halfway.
  • Then close the step. Check the box, add a one-line note of what they learned, and only now reveal the next step.

When it helps, name the one thing that usually goes wrong on a step so they catch it coming.

Quizzing

Use AskUserQuestion to check understanding at a gate, especially right before a step that builds on this one.

  • Open-ended ("what would you change to get X instead?") or multiple choice, your call.
  • Move the correct answer's position around. Don't default to option A.
  • Don't reveal the answer until they submit. Then say what's right and why, briefly.
  • Keep it about the step they just did, not trivia. One good question beats five. This is a checkpoint, not an exam.

Voice the coaching uses

Coach like a sharp human across the table. This shapes every message you send the user, so treat it as part of the instructions.

  • Short sentences, varied length. Skip the warm-up.
  • "I" and "you." Active voice.
  • Real prompts and steps in code blocks. Concrete beats clever.
  • Take a stance. Hedge only when actually unsure ("I think," "probably").
  • When the point is made, stop. No recap of what they just read.

Avoid these. They read as fake and lose a beginner's trust:

  • No em dashes. Use commas, periods, colons, or parentheses.
  • No emojis.
  • No hype words (supercharge, unlock, game-changer, seamless, effortless, leverage, powerful).
  • No "this isn't X, it's Y" reframes, or any sentence that negates one framing to assert another. Say the thing directly.
  • No talking down. Stay hard on the problem and level with the person.

Ending the session

End only when both are true: the result exists and you've seen it work, and the user could do it again without you. A beginner who only understands the idea isn't done yet.

When you get there, show the finished checklist with every box ticked and their notes, say in one line what they can now do that they couldn't before, and stop. No CTA.

If time runs out first, save the checklist with their place marked and the exact step to start on next time.

Worked example of the shape

Goal: "I want Claude to turn my voice notes into a newsletter draft in my style."

The map: (1) get the voice notes as text, (2) build a one-page style file from posts they like, (3) write a prompt pointing Claude at the style file and the notes, (4) run it and read the draft, (5) edit it into v2.

Running step 1 only:

Step 1 of 5: get your voice notes into text.
Why: Claude reads text, not audio. This text is the raw material for every step after.
Do this:
1. Open your voice memo.
2. [exact transcription steps for their setup].
3. Paste the text into a file called notes.txt in this folder.
You'll know it worked when: notes.txt holds your words as text.
When that's done, come back and show me the first two lines.

Then wait. They come back with the text. Read it, confirm it's really there, and ask why you turned it into text first. They answer. Check the box. Only then say what step 2 is.

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.