Be Specific Instead Of Vague

Be Specific Instead of Vague (The Precision Protocol) (AI Skill)

Overview

The quality of an AI's response is directly proportional to the constraints provided in the prompt. Vague prompts ("Write a blog post about productivity", "Fix my code") trigger the broadest, most generic probability distribution in the LLM, producing bland, predictable text.

The Precision Protocol uses the 5-Variable Specificity Formula to anchor the model to your exact intended context, eliminating 90% of back-and-forth revisions.

The 5-Variable Specificity Formula

┌─────────────────────────────────────────────────────────────┐
│                 The 5-Variable Formula                      │
│                                                             │
│  [ ROLE ]     ──► Who is the AI speaking as?                │
│  [ AUDIENCE ] ──► Who is reading / receiving this?          │
│  [ GOAL ]     ──► What exact outcome must be achieved?      │
│  [ BOUNDARY ] ──► Length, format, and banned elements?      │
│  [ CONTEXT ]  ──► Key facts, data points, or constraints?   │
└─────────────────────────────────────────────────────────────┘

Real-World Transformations: Vague vs. Specific

Example 1: Email Request to a Colleague

Vague (Low Precision)

❌ "Write an email asking for the sales numbers."

Result: AI writes a 4-paragraph corporate letter starting with "I hope this email finds you well in these busy times..."

Specific (Precision Formula Applied)

✅ "Write a 2-sentence Slack message to our sales lead (Dave) asking for the updated Q2 EMEA revenue spreadsheet by 3 PM today for the board prep meeting. Friendly but urgent tone."

Result: "Hey Dave, could you share the updated Q2 EMEA revenue spreadsheet by 3 PM today? Putting the final slides together for tomorrow's board prep."

Example 2: Code Refactoring

Vague

❌ "Make this Python script faster."

Result: AI adds comments, renames variables, or suggests generic multiprocessing libraries that don't fit the workload.

Specific

✅ "Refactor this Python 3.12 data processing function to eliminate the $O(N^2)$ nested loop. Use a hash map lookup (dict). Keep memory footprint under 50MB and preserve all existing type hints and docstrings."

Example 3: Marketing Copy

Vague

❌ "Give me headline ideas for a fitness app."

Result: "Get fit today!", "Transform your body!", "Your fitness journey starts now!"

Specific

✅ "Write 5 headline options for a high-intensity kettlebell workout app targeted at busy software engineers who have only 20 minutes a day. Highlight strength and posture improvement. Under 10 words each. No exclamation marks."

Result:

1. Fix developer posture in 20 minutes a day
2. Build real strength between Zoom meetings
3. Zero commute kettlebell training for desk athletes
4. 20 minutes of iron to counter 8 hours of sitting
5. Heavy kettlebells, zero fluff, done in 20 minutes

Quick-Constraint Injection Table

When you are in a rush, append any 2 of these constraints to instantly sharpen your prompt: