Keep this open

The prompt cheat sheet

The whole method on one page: the structure of a strong prompt, reusable templates, and every term in plain language. Bookmark it. When you are ready to actually get good, take the free class.

Universal prompt skeleton

You are a {{role}}. Task: {{the single clear objective}} Context: - Audience: {{who reads this}} - Why it matters: {{motivation}} - Source material: {{wrapped in tags}} Constraints: - {{hard limits, length, scope, tone}} - Do X (state positively, not "don't do Y") Output format: - {{exact shape, prose / JSON schema / sections}} Before finishing, verify the result against: {{named criteria}}

Decision quick-reference

Simple task → zero-shot, Task + Format only.
Specific format/tone → add a few clean examples (fewer on reasoning models).
Hard reasoning, capable model → goal + constraints, raise effort, test before adding CoT.
Must be correct → named verification step (and self-consistency if critical).
Output feeds a system → structured output / schema.
Tools / multi-step → lean system prompt, autonomy default, safety rails.
Untrusted content → injection-defence wrapper (the Safety module).

When a prompt fails, in order

Is the task unambiguous? Add the missing component.
Did you give the why?
Is the key instruction first or last, not buried?
On a reasoner, try removing scaffolding, not adding it.
Still failing? Iterate against a test set, one change at a time.

Three reusable templates, copy or launch

Minimal reasoning-model

Task: [single objective]

Constraints:
- [scope, tone, exclusions]
- [success criteria]

Output:
[exact sections / schema]

Before finishing, check the result against [named criteria].

Retrieval-grounded

<documents>
[paste your source text here]
</documents>

Question: [your question]

1. Extract the quotes relevant to the question.
2. Answer using only those quotes.
3. Cite each source inline.
4. If the evidence is thin or missing, say so plainly.

Prompt-evaluation

Grade the prompt below from 1-4 on: objective clarity, context sufficiency, output contract, robustness to ambiguity, and safety + evidence.

Return JSON: {"scores": {...}, "highest_risk": "...", "one_best_improvement": "..."}

Prompt to grade:
"""
[paste the prompt you want graded]
"""

Reference

Glossary

Acronyms are defined on first use in the modules; this is the quick reference.

Zero-shot	Prompting with instructions only, no examples.
Few-shot	Providing example input→output pairs to steer behavior.
CoT (Chain-of-Thought)	Eliciting step-by-step reasoning before the final answer.
Self-consistency	Sampling multiple reasoning paths; taking the majority answer.
ReAct	Reason→Act→Observe loop; the basis of tool-using agents.
ToT (Tree-of-Thoughts)	Branching, scoring and pruning multiple reasoning paths.
Context engineering	Curating the useful set of tokens across a task.
Structured outputs	Constrained decoding that yields schema-valid output.
Effort / thinking budget	A control for how much a model reasons before answering.
System / developer message	The authoritative instruction defining role and rules.
RAG (Retrieval-Augmented Generation)	Grounding answers in fetched documents.
Prompt injection	Untrusted content trying to hijack the model's instructions.
LLM-as-judge	Using a model to score outputs against criteria at scale.