The Meta-Prompt Move

Use it for

Let the AI Think: The 80% Humanization Experiment

The prompt

# The Meta-Prompt Move

Give Claude the goal instead of the method — have it design the optimal prompting strategy, interaction structure, and output format before producing any content.

MY GOAL: $ARGUMENTS

MY CONSTRAINTS: [length / format / audience / deadline / tools available / "none" if flexible / "you decide"]

STEP 1 — GOAL DECOMPOSITION: What is the actual outcome needed? Context for use? Implicit quality criteria? Max 3 clarifying questions.
STEP 2 — STRATEGY DESIGN: Approach architecture (one pass or multiple?) → Role selection (specific expert type) → Quality criteria (failure modes to avoid) → Output format (structure, length, inclusions/exclusions)
STEP 3 — INFORMATION GAP CHECK: Must-have info (blocking) vs. nice-to-have (assumptions noted). Ask for must-haves only.
STEP 4 — EXECUTE: Produce output following the designed strategy. Don't revert to defaults mid-execution.
STEP 5 — SELF-AUDIT: Did I follow my own strategy? Does output meet my quality criteria? Which failure modes crept in? Revise before delivering.

CONSTRAINTS: Strategy visible to user (teaching value). Don't over-engineer simple goals. Adapt depth to complexity.

Full prompt on recap page.

Worked example

No worked example recorded yet.