agentleFS
Sign inStart free
seven-hills-softwareseven-hills-software/skillwright-skills/prompt-writes-prompt/SKILL.md

prompt-writes-prompt

Skill0 starsChanged 5 months agoMITView on GitHub

When the user describes a task vaguely, first draft the best possible prompt for that task, show it, then execute it.

The file

We split this skill into the rules it gives an agent

3 rules. About 0 are common, 3 are distinctive, and 0 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈3

    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:

    After showing the drafted prompt:

    line 42
  2. Distinctive:

    If the user says run it, immediately execute the prompt you just wrote — do not ask for additional input the prompt already covers.

    line 46
  3. Distinctive:

    If the user adjusts, integrate the changes and run the revised version.

    line 48

≈ “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

Prompt Writes Prompt

Pattern: instead of executing a vague request directly, write the well-formed prompt for that request first, show it to the user, then run it.

When this skill is active

The user says any of:

  • "Help me with X" (where X is one sentence)
  • "Write me a prompt for Y"
  • "How should I prompt you to do Z"
  • The request is under-specified and the user has not yet defined role, format, constraints, or success criteria

The two-step pattern

Step 1 — Draft the prompt

Produce a complete prompt for the user's task. Include:

  1. Role. Who should the assistant be acting as for this task?
  2. Context. What background does the assistant need that the user is assuming?
  3. Inputs. What the user will provide.
  4. Output format. Structure, length, tone, voice.
  5. Constraints. What to avoid, what to never do, what is non-negotiable.
  6. Success criteria. How the user will judge if the output is good.

Format it as a single prompt block, ready to copy. Be specific. Avoid filler.

Step 2 — Confirm, then execute

After showing the drafted prompt:

"Run this as-is, or do you want to adjust the role / constraints / format first?"

If the user says run it, immediately execute the prompt you just wrote — do not ask for additional input the prompt already covers.

If the user adjusts, integrate the changes and run the revised version.

Why this pattern works

  • Forces the assistant to commit to a frame before generating output
  • Makes the framing visible and editable to the user
  • Catches misunderstandings on the prompt, where revisions are cheap, not on the 800-word output
  • Teaches the user to write better prompts by example

When NOT to use this pattern

  • The user's request is already a complete prompt
  • The task is mechanical (rename, format, lint, search)
  • The user has explicitly opted out ("just do it, don't draft")
  • The cost of the drafted prompt exceeds the cost of the task itself

Quality bar for drafted prompts

  • The prompt would still make sense if shown to a different assistant
  • It does not reference shared context the user has not stated
  • It produces a deterministic-ish frame, not a vague invitation
  • It is under 250 words unless the task genuinely requires more

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