For anyone leveling up their prompting · LEARN

Practice prompts like a skill.Graded drills that stick.

Write a prompt for a real task, score it against the six-criterion RCTFCV rubric, and let spaced repetition schedule the next review as you improve. It grades the craft — the structure of what you wrote — not the model's luck.

Get the Lab — $69one-time · instant download · yours to keep
Five deliverables · runnable
Drill engine (Python)
runnable
Drill Lab workbook (.xlsx)
reproduces it
The RCTFCV Rubric
playbook
Deliberate Practice + Spacing
playbook
Sample drill deck
worked example
Works alongside
Operator Prompt Vault · Anti-Slop System · Blueprint Library
01.The Problem

Judging a prompt by whether the output looked good trains luck, not skill.

Vibe ≠ skill

A good output can come from a sloppy prompt and a lucky model. Grade the prompt's structure and you improve what you control.

One-and-done fades

Read a prompting guide once and it's gone in a week. Skill needs spaced reps with feedback — the conditions practice grows under.

The skipped habit

Most prompts never ask the model to check its own work. That's the single habit that separates reliable from lucky.

02.See It Work

Edit the ratings — a high total still can't be STRONG without the Verification-ask.

Drill lab · live
DRILLING
STRONG 2 · WORKABLE 1 · WEAK 1
mean 74 (context)

A drill is held off STRONG — its Verification-ask is below the floor (needs 4+). A strong total never rescues a missing ask-the-model-to-check-itself.

Cold outreach email to a CFO
100STRONG
box 2 → 3 · review in 4 sessions
Summarize a support thread
75STRONG
box 1 → 2 · review in 2 sessions
Competitor teardown (high total no V-ask)
85WORKABLEv-ask
box 3 → 3 · review in 4 sessions
Blog outline from a transcript
34WEAKv-ask
box 1 → 1 · review in 1 session

Grades prompt craft against the RCTFCV rubric — the structure of what you wrote, not the model's output. STRONG requires Verification-ask of 4 or more. No tool makes you an expert in N days.

03.The Engine, Run

The same scoring runs as a Python engine and a workbook that reproduces it exactly.

Verbatim output from the included sample deck. The competitor-teardown drill scores 85 — but it caps at WORKABLE because its Verification-ask is below the floor. STRONG drills promote their Leitner box; the WEAK one resets to box 1.

==================================================================
  PROMPT PRACTICE LAB  -  graded drills with spaced repetition
  Practice session as of 2026-06-23
==================================================================

  SESSION:  DRILLING
  Drills: 4   STRONG 2  WORKABLE 1  WEAK 1   Mean score: 74 (context only)
  ! GATE: a drill is held off STRONG because its Verification-ask is below floor.

  PER DRILL
  --------------------------------------------------------------
    Cold outreach email to a CFO   100  STRONG     box 2->3 (review in 4 sess)
    Summarize a support thread      75  STRONG     box 1->2 (review in 2 sess)
    Competitor teardown (high total no V-ask)  85  WORKABLE   box 3->3 (review in 4 sess)  <-- V-ASK GATE
    Blog outline from a transcript  34  WEAK       box 1->1 (review in 1 sess)  <-- V-ASK GATE

  Grades prompt craft against the RCTFCV rubric - not the model's
  output. STRONG needs Verification-ask >= 4: ask the model to
  check itself, or you cannot be a strong prompter.
==================================================================
04.The Standard

Three principles keep the practice honest.

Grade the prompt, not the output

The rubric scores the structure of what you wrote — role, context, task, format, constraints, verification-ask — not whether the model happened to nail it.

One gate you can't score around

A drill caps at WORKABLE if Verification-ask is below 4, however high the total. The cap only lowers a would-be STRONG; it never lifts a WEAK.

Spacing, not cramming

STRONG promotes a card to a longer interval; WEAK resets it to frequent review. Intervals are counted in sessions, so the schedule never drifts.

05.What It Is — And Isn't

A skill-builder for the person writing the prompt. Not a CI tool, not a model grader.

It is
  • A rubric-graded practice loop for your own prompt-writing skill.
  • A spaced-repetition schedule that fades the scaffolding as you improve.
  • A runnable engine + a workbook that reproduces it, run on your own drills.
It isn't
  • A CI prompt-diff tool (that's the Prompt Regression Lab) or an eval framework.
  • A grader of the model's output or of your subject knowledge.
  • A promise to make you an expert in N days — skill takes real reps.
Scope

The Lab grades prompt craft against a rubric — the structure of what you wrote — not the quality of the model's output and not your knowledge of any subject. No tool makes you an expert in a fixed number of days. Not legal, medical, or professional advice.

06.Who It's For

Anyone who wants prompting to be a reliable skill, not a lucky streak.

For you if
  • You use AI daily and want your prompts to get reliably better.
  • You're training a team and want a shared rubric and a practice cadence.
  • You learn by doing — real drills with immediate, structured feedback.
Not for you if
  • You need CI prompt-regression testing — use the Prompt Regression Lab.
  • You want done-for-you prompts to paste — grab the free Operator Prompt Vault.
  • You want a tool that grades the model's answers for you.
07.Pairs Well With

Build the skill here, then put it to work.

Anti-Slop Content System

Once your prompts are STRONG, keep the output from sounding like everyone else's AI.

View
AI Data Analysis Starter Kit

Apply the verification-ask habit to prompts that run code on your own uploads.

View
Claude Project Blueprint Library

Drop your best drills into a reusable Claude Project for each business function.

View
08.Common Questions

Straight answers on what it grades, the gate, and how the spacing works.

It grades the structure of your prompt against an explicit six-criterion rubric (RCTFCV: Role, Context, Task, Format, Constraints, Verification-ask), not the quality of the model's output and not your knowledge of the subject. Each criterion scores 0-5; weights sum to 100; the weighted total bands the drill STRONG (70+), WORKABLE (40-69), or WEAK (under 40). You improve the one thing you control — how you write the prompt.

Stop guessing at prompts.
Practice them like a skill.

One purchase, lifetime access, 12 months of updates. $69, once.

Get the Lab — $69

Sold by RedHub AI LLC · Secured by Stripe · redhub.ai