prompt-improver
Skill0 starsChanged 7 days agoMITView on GitHub
Turns a rough request or weak prompt into a strong one. Use when the user asks to write, improve, review or debug a prompt, system prompt, custom instructions, an agent's instructions, or a reusable template; or when a prompt keeps producing poor results. Checks the six elements (role, context, instruction, format, constraints, examples), predicts likely failures, and returns a tested, leaner prompt.
12 other projects use a version of this.
Adopted by
Most recently taken up by
- KalastajaM/Cluidetheir own version, first seen Nov 8, 2026
- zedarvates/frugal-skillstheir own version, first seen Nov 8, 2026
- No1Hoo/prompt-improver-skilltheir own version, first seen Sep 14, 2026
- Amery2010/skillstheir own version, first seen Sep 9, 2026
- davekilleen/Dextheir own version, first seen Aug 31, 2026
- d3v0psdan/claude-skillstheir own version, first seen Jun 19, 2026
Other versions12 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 12
- severity1/claude-code-prompt-improver★ 1.9ksame name
- davekilleen/Dex★ 509same name
- happycapy-ai/Happycapy-skills★ 137same name
- ndpvt-web/prompt-improver★ 86same name
- dazuck/operator-skills★ 17same name
- Amery2010/skills★ 4same name
- KalastajaM/Cluide★ 2same name
- dazuck/claude-code-skills★ 1same name
- zedarvates/frugal-skills★ 1same name
- d3v0psdan/claude-skills★ 0same name
2 more not listed
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 repository11 other files it gives its agents.
- drop-in/AGENTS.mdAGENTS.md
- llms.txtllms.txt
- clarify-firstSkill
- coverage-checkSkill
- fresh-eyes-reviewSkill
- fresh-start-handoffSkill
- grounded-answerSkill
- honest-critiqueSkill
- learning-tutorSkill
- task-specSkill
- verify-claimsSkill
The file
We split this skill into the rules it gives an agent
3 rules. 0 are common, 3 are distinctive, and 0 are very specific to this project.
0
Common
Most of the 10 edited versions we read keep these.
3
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
- Distinctive: line 18
Constraints: only the rules that truly matter (about five or fewer), each with a reason
- Distinctive: line 20
Role: only for tone or perspective.
- Distinctive: line 23
Trim: Remove ALL-CAPS emphasis, "think step by step" for reasoning models, threats, tips, flattery of the model, and duplicate rules.
“Common” is measured: kept by at least 6 in 10 of the 10 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.
Files
Prompt improver
Good prompts are not long; they leave little for the model to guess. Improve prompts by filling the gaps that cause failures and deleting what doesn't help.
Steps
- Understand the job. Identify the task, audience, where the output will be used, and what "done" looks like. If these are missing and matter, ask (at most three questions).
- Check the six elements and note what's missing or weak:
- Context (most important): purpose, audience, situation, source material
- Instruction: the exact action and the done criteria
- Format: structure, length, tone
- Constraints: only the rules that truly matter (about five or fewer), each with a reason
- Examples: 2–5 varied examples if style or format is hard to describe (try none first for reasoning models)
- Role: only for tone or perspective; it doesn't improve accuracy
- Predict failures. List the most likely ways the prompt could fail: misunderstanding (ambiguity), omission (too many requirements, buried in a long input), fabrication (no source, no permission to say "I don't know"), sycophancy (asking for opinions on "my" work).
- Rewrite. Return the improved prompt in a code block. Put long documents first and the question last; separate parts with consistent tags or headings; add "If something isn't in the material, say so" where facts matter. Use clearly marked placeholders for facts you don't have (policies, prices, limits, names), and keep invented example values out of the prompt: the model will treat them as real.
- Trim. Remove ALL-CAPS emphasis, "think step by step" for reasoning models, threats, tips, flattery of the model, and duplicate rules.
- Explain and test. Give a short table of what changed and why, and suggest 3–5 test inputs (including an edge case) to compare old vs. new.
Evidence
- Prompt components and ordering: Anthropic prompting best practices — https://platform.claude.com/docs/en/build-with-claude/prompt-engineering/claude-prompting-best-practices ; OpenAI prompt engineering guide — https://developers.openai.com/api/docs/guides/prompt-engineering
- Format instructions measurably help; tips, threats and expert personas don't reliably help: Wharton Prompting Science Reports — https://arxiv.org/abs/2503.04818 , https://arxiv.org/abs/2508.00614 ; personas: https://arxiv.org/abs/2311.10054
- "Think step by step" adds little for reasoning models (2025): https://arxiv.org/abs/2506.07142
- More in the guide: https://github.com/ji-hun-git/how-to-use-ai/blob/main/docs/02-prompt-anatomy.md
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
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.

