The 7 Deadly Sins of Prompt Writing (and How to Overcome Them) - Zroam Tools

Large language models (LLMs) are powerful tools, but they’re only as good as the prompts you give them. Writing effective prompts is an art and a science. Falling prey to common pitfalls can lead to underwhelming or even inaccurate results. This article explores the 7 deadly sins of prompt writing and offers practical advice on how to overcome them.

Sin #1: Vagueness (Lust for Ambiguity)

The LLM is not a mind reader. Vague prompts yield vague responses. Be specific and define your expectations clearly.

Bad Prompt: “Write a story.”

Good Prompt: “Write a short science fiction story set on Mars about a botanist who discovers an alien artifact that can help her grow plants.”

The improved prompt provides specific details about genre, setting, main character, and plot element.

Sin #2: Lack of Context (Gluttony for Information)

LLMs need context to understand your intent. Don’t assume the model knows what you’re talking about. Provide background information and relevant details.

Bad Prompt: “Explain the issue.”

Good Prompt: “Explain the issue of climate change, including its causes, potential impacts, and possible solutions.”

The improved prompt clarifies the topic of the “issue.”

Sin #3: Absence of Structure (Greed for Power)

Structure your prompts logically. Break down complex requests into smaller, manageable steps. Use keywords and phrases that guide the LLM’s output.

Bad Prompt: “Compare these things.”

Good Prompt: “Compare and contrast the advantages and disadvantages of using Python versus Java for web development. Consider factors like performance, ease of use, community support, and available libraries.”

The improved prompt explicitly states the desired comparison points.

Sin #4: Ignoring Tone and Style (Sloth of Communication)

Specify the desired tone and style of the output. Do you want it to be formal, informal, humorous, or technical?

Bad Prompt: “Write a review.”

Good Prompt: “Write a humorous and slightly sarcastic review of the movie ‘Plan 9 from Outer Space’.”

The improved prompt sets the tone and provides the subject of the review.

Sin #5: Neglecting Constraints (Wrath of the Unlimited)

Define constraints such as length, format, or specific keywords that must be included or excluded. Without constraints, the model might wander off-topic or produce output that doesn’t meet your needs.

Bad Prompt: “Summarize this article.”

Good Prompt: “Summarize this article in three bullet points, each no more than 25 words long. The summary should focus on the main arguments presented in the article.”

The improved prompt specifies the desired length, format, and focus of the summary.

Sin #6: Lack of Examples (Envy of Expertise)

Providing examples can significantly improve the quality of the output. Show the LLM what you’re looking for by demonstrating the desired style and format.

Bad Prompt: “Write a poem.”

Good Prompt: “Write a haiku about autumn, following this example: Falling leaves descend, / Crimson, gold, and amber hues, / Winter’s breath draws near.”

The improved prompt includes a haiku example to guide the LLM’s output.

Sin #7: Insufficient Iteration (Pride of First Drafts)

Don’t expect perfect results on the first try. Prompt engineering is an iterative process. Review the LLM’s output and refine your prompts to get closer to your desired outcome. Experiment with different phrasing and approaches.

By avoiding these 7 deadly sins, you can significantly improve the quality and relevance of the responses you receive from LLMs. Happy prompting!