Hire an AI prompt and skill improvement tester

Everything it can hand back

Your task sets the actual outputs. The specialist picks the methods that cover them.

• Scores before and after, with the rounds used out of four
• The new version as a copy beside your original
• A log of every change, kept or undone, with why
• A list of what still fails and the next change to try

How it does the job well

• Clarify the purpose, audience and limits, such as length, format and deadline, first.
• Use the user's materials and voice. Ask for samples when tone matters.
• Be specific and concrete. Cut what does not serve the purpose.
• Say what you are unsure of and where to check it.

Mistakes it avoids

• Tuning to the test cases until the prompt only works on them. Keep 3 to 5 cases aside that you do not tune against, and run them on the final version only.
• Changing three things in one round, so nobody knows which one helped.
• Adding more rules to fix every miss. A longer prompt or skill is often a worse one.
• Calling a skill safe because a checker found nothing. Say what was and was not checked.
• Generic output that ignores the user's details.
• Inventing facts, quotes or sources.

What a good result looks like

An example, not your task: A prompt gives weak or uneven answers and you want fixes that are tested, not guessed.

• RESULT. One line: passed or not passed, rounds used out of four, and the score before and after.
• NEW VERSION. The tuned prompt or skill folder, saved as a copy beside your original.
• WHAT CHANGED. One line per kept change, with the check it fixed and the score move.
• UNDONE. Each change that made things worse, and why it was undone.
• STILL FAILING. Checks still below your pass line, and the next change to try.
• SAFETY. The safety check result for a skill, or not run for a prompt.

How the work is checked

• The same test cases were used in every round.
• Each round changed one thing, and its effect is in log.md.
• No kept change broke a case that passed before.
• The original is untouched.
• States which checks passed, failed or could not run, and names missing inputs.

These checks have not been run against a model yet. Your AI uses them to check its own work.

Capabilities

• Set the pass line. Agrees what counts as done, and the cap of four rounds, before any change is made.
• Check the test cases. Makes sure there are enough varied cases, and that you confirmed any it drafted.
• Test the baseline. Scores the original. If it already passes, it stops and leaves it alone.
• Loop up to four rounds. Changes one thing, tests again, keeps the change only if the score rose and no passing case broke, and undoes it otherwise.
• Check skills for safety. Runs a safety check on the final skill. A serious finding ends the run with a do-not-use note.
• Hand back the copy. Gives you the new version, the log and what still fails, and stops for your yes before anything is shared or replaced.

How the role works

1. Checks your task, context and available tools.
2. Picks the methods needed and prepares a work plan.
3. Carries out supported tasks, with each method's limits and review gates.
4. Checks the real deliverables and corrects supported faults.
5. Hands back files, evidence and unresolved gaps.

Your specialist's working setup

Download the base specialist without a task form or account. README.md and SETUP.md explain how to use the folder. SOUL.md defines its scope. INTAKE.md asks missing questions inside your AI. PLAYBOOK.md, SKILLS.md, BEST_PRACTICES.md, CHECKS.md and HANDOFF.md guide the work and its checked deliverables.

Complete associated skills, supporting files and licences are included. Choose only the methods needed for the actual task. Keep context, working files and accepted feedback in your AI project.

Use this specialist through MCP or the API

Call prepare_specialist with role_id prompt-and-skill-tuner only. No task is required to load it. Your AI asks what job you need done, what you want back and which relevant context it can use.

For task-led hiring, call find_specialist, explain the suggested specialists and let the user accept or change the team.

Complete ZIP: Download base specialist. API: GET /api/v1/specialists/prompt-and-skill-tuner/pack. The host AI supplies its own model, tools and execution; a pack does not connect accounts or run a task.

Inputs and helpful context

The specialist starts from the task brief. Only active methods require their inputs.

For the primary method, when used

• The prompt text, or the skill folder with its SKILL.md
• 10 to 20 real test cases, each with what a good answer looks like
• Your pass line, if you have one (optional)
• Examples of answers you did not like (optional)
• A paid model or API key only if you choose to test that way. By default it tests by hand in the chat, and it runs no script that comes with a skill without your yes

Helpful context (optional)

• Past answers you liked and answers you did not
• The input data the prompt will meet in real use

Tools and access

• Claude Code: It can read the skill folder, keep versions in the tuner folder and run each round.
• A spreadsheet: A simple sheet of cases and pass or fail marks makes each round easy to compare.
• Any good chat AI: Claude, ChatGPT or Gemini can all write and review a SKILL.md as text.
• Claude Code: It reads skill folders directly, so you can test that the skill fires on its own.

Functions and deliverables

• Prompt and skill improvement tester: Test a prompt or skill against agreed cases. Try one change at a time and keep fixes that avoid regressions.
  • Scores before and after, with the rounds used out of four
  • The new version as a copy beside your original
  • A log of every change, kept or undone, with why
  • A list of what still fails and the next change to try
• Prompt tester and tuner: Test a prompt against examples. Get scores for each version and a revision to review.
  • Baseline score
  • Scores for each new version
  • An improved prompt, with a pass or fail result per test case from your own runs
  • Token and cost estimate
  Source: Alireza Rezvani.
• Prompt fixer: Fix a prompt that gives vague or uneven answers. Paste it in with one bad reply and get a tighter version that works more reliably.
  • Diagnosis of what is wrong
  • Rewritten prompt
  • Why each change helps
  • Tips to test it
  Source: sickn33.
• Marketing prompt tester: Compare prompt versions on shared test cases. Get scores, flagged failures and a saved version history.
  • Scores for each prompt on each case
  • A named winner with evidence
  • A saved version history
  • A change log between versions
  Source: Alireza Rezvani.
• Skill file validator: Share a SKILL.md. Get structure, wording and naming problems to fix before you share it.
  • A list of problems by category
  • Points where the file does well
  • A quality score
  • Fixes for the issues found
  Source: sickn33.
• Skill quality tester: Give it a skill folder. Get structure and script checks, a quality score and a ranked fix list.
  • A structure check result
  • Script test results
  • A quality score and grade
  • An improvement road map
  • A security score
  Source: Alireza Rezvani.
• Skill trigger auditor: Share your skills. Get a grade for each on when it fires, and which skills block each other.
  • An A to F grade for every skill
  • A ranked list of which skill fires for your sample request
  • A list of skills that block each other
  • Suggested "Use when" lines for weak descriptions
  Source: sickn33.
• Skill writer and reviewer: Write an AI skill from a job description, or review one you have. Get a clear trigger line, safety notes and a cleaner SKILL.md.
  • A trigger line that says what, when and how it differs
  • Structure notes on length and what to move out
  • Safety notes on code, web access and file changes
  • A rewritten SKILL.md
  • Three test prompts that should fire it and two that should not
  Source: sickn33.
• Skill safety checker: Share a skill or bundle before you add it. Get risky code, data leaks and hidden payloads.
  • Risky patterns found, grouped by threat
  • What each finding could do
  • Platform-specific notes
  • A summary of what to remove or ask about
  Source: sickn33.
• me.md builder: Answer a few questions, or let your AI use what it remembers. Get one private file that tells any AI who you are.
  • One me.md file with up to ten dated sections
  • A list of facts taken from memory, for you to check
  • A short note of what it removed for privacy
  • The two next steps, to save it and set it up

Tasks you can use it for

• A prompt gives weak or uneven answers and you want fixes that are tested, not guessed.
• You wrote a skill and want it checked, fixed and checked again before you use it.
• You are about to share a skill, or add one from someone else, and want a safety check first.

Limits to check before starting

• You want a one-off piece of writing: ask your AI directly, no tuning needed.
• Your prompt already works well: leave it alone, or fix one weak spot by hand with a quick edit.
• You only want a single check of one prompt: run it through the Prompt tester and tuner instead.

Best way to use this role

Give it one clear outcome, the actual files and a definition of done. Keep the work in one AI project so it can reuse checked context and earlier files.

• Start a chat with the agent. In Claude Code, add it as a skill or save it in .claude/agents, then ask it to tune your prompt or skill.
• Give it the prompt or the skill folder, and your test cases. If you have no test cases, it helps you build them first.
• Confirm the pass line it proposes, or give your own.
• Read the hand-back. Your original is untouched. Copy over the new version yourself, or say yes and it saves it where you say.
• Run it again whenever you change the prompt or skill.

Read it as data

Read the complete role setup as JSON. Connect your AI to the AI Blocks MCP library and read aiblocks://roles/prompt-and-skill-tuner for its role instructions and pinned sources.

This is a prepared AI role. No task has run. Your AI carries out authorized work using its available tools and reports the actual deliverables and checks.