Instruction Clarity
Tier: Foundational · Task types: All · Difficulty: ⭐☆☆☆☆
Table of contents
Definition
Instruction clarity is the practice of writing prompts that are specific, unambiguous, and complete — eliminating vague verbs, underspecified targets, and implicit assumptions. It is less a discrete technique and more a foundational discipline: the quality ceiling of any other technique is set by the clarity of its instructions.
The test for instruction clarity: could two different people read your prompt and produce the same output? If not, the model will make arbitrary choices to fill the gaps.
Use Cases
- Repeatable task pipelines — when the same prompt will run on hundreds of inputs, ambiguity compounds across every result
- Multi-step instructions — decomposing complex tasks into clearly ordered, numbered steps prevents the model from re-ordering or conflating steps
- Cross-team prompts — when prompts are shared between team members, clarity prevents “it works differently for me” problems
Weak Prompt
Make this email better.
[Original email pasted here]
Weak Prompt Output (Gemini)
[Rewrite with minor phrasing improvements and a slightly warmer closing]
Assessment: “Better” is undefined. The model made conservative style edits. The user may have wanted structural changes, tone shifts, length reduction, or a complete rewrite — the model had no way to know.
Improved Prompt
Rewrite the following email to achieve these specific goals:
1. Reduce length by approximately 40% — cut filler sentences
2. Make the ask in paragraph 3 explicit and move it to the opening
3. Adopt a direct, professional tone — remove hedge phrases like "I just wanted to" and "sorry to bother you"
4. Keep all factual details unchanged
Return only the rewritten email. Do not explain what you changed.
[Original email pasted here]
Improved Prompt Output (Gemini)
[Rewrite that leads with the ask, is measurably shorter, removes hedging language, and preserves factual content]
Assessment: Every instruction is unambiguous and independently verifiable. The model has a clear quality target for each dimension.
Verdict
Instruction clarity underlies the effectiveness of every other technique in this playbook. Before reaching for chain-of-thought or few-shot examples, check whether the basic instruction is specific enough. Most prompt failures are instruction clarity failures in disguise — the model chose an interpretation you didn’t intend.
Common Pitfalls
1. Using vague evaluation verbs “Improve”, “enhance”, “fix”, “make better” are all undefined. Replace with: “reduce to 200 words”, “remove passive voice”, “add a call-to-action in the last sentence.”
2. Implicit assumptions about format If you want bullet points, say so. If you want a single paragraph, say so. The model’s default format choice may not match your expectation.
3. Burying the most important instruction Models attend more strongly to instructions at the beginning and end of prompts. Put your most critical requirements first, not last.