prompt-engineering

Name: prompt-engineering
Author: ancoleman

62installs59forks

How to install prompt-engineering?

ancoleman/ai-design-components/prompt-engineering

$npx -y skills add ancoleman/ai-design-components --skill prompt-engineering

Installs into the current project.

Summary

Engineer effective LLM prompts using zero-shot, few-shot, chain-of-thought, and structured output techniques. Use when building LLM applications requiring reliable outputs, implementing RAG systems, creating AI agents, or optimizing prompt quality and cost. Covers OpenAI, Anthrop

Files · 1

View on GitHub

SKILL.md

1	# Prompt Engineering
2
3	Design and optimize prompts for large language models (LLMs) to achieve reliable, high-quality outputs across diverse tasks.
4
5	## Purpose
6
7	This skill provides systematic techniques for crafting prompts that consistently elicit desired behaviors from LLMs. Rather than trial-and-error prompt iteration, apply proven patterns (zero-shot, few-shot, chain-of-thought, structured outputs) to improve accuracy, reduce costs, and build production-ready LLM applications. Covers multi-model deployment (OpenAI GPT, Anthropic Claude, Google Gemini, open-source models) with Python and TypeScript examples.
8
9	## When to Use This Skill
10
11	Trigger this skill when:
12	- Building LLM-powered applications requiring consistent outputs
13	- Model outputs are unreliable, inconsistent, or hallucinating
14	- Need structured data (JSON) from natural language inputs
15	- Implementing multi-step reasoning tasks (math, logic, analysis)
16	- Creating AI agents that use tools and external APIs
17	- Optimizing prompt costs or latency in production systems
18	- Migrating prompts across different model providers
19	- Establishing prompt versioning and testing workflows
20
21	Common requests:
22	- "How do I make Claude/GPT follow instructions reliably?"
23	- "My JSON parsing keeps failing - how to get valid outputs?"
24	- "Need to build a RAG system for question-answering"
25	- "How to reduce hallucination in model responses?"
26	- "What's the best way to implement multi-step workflows?"
27
28	## Quick Start
29
30	Zero-Shot Prompt (Python + OpenAI):
31	```python
32	from openai import OpenAI
33	client = OpenAI()
34
35	response = client.chat.completions.create(
36	model="gpt-4",
37	messages=[
38	{"role": "system", "content": "You are a helpful assistant."},
39	{"role": "user", "content": "Summarize this article in 3 sentences: [text]"}
40	],
41	temperature=0 # Deterministic output
42	)
43	print(response.choices[0].message.content)
44	```
45
46	Structured Output (TypeScript + Vercel AI SDK):
47	```typescript
48	import { generateObject } from 'ai';
49	import { openai } from '@ai-sdk/openai';
50	import { z } from 'zod';
51
52	const schema = z.object({
53	name: z.string(),
54	sentiment: z.enum(['positive', 'negative', 'neutral']),
55	});
56
57	const { object } = await generateObject({
58	model: openai('gpt-4'),
59	schema,
60	prompt: 'Extract sentiment from: "This product is amazing!"',
61	});
62	```
63
64	## Prompting Technique Decision Framework
65
66	Choose the right technique based on task requirements:
67
68	\| Goal \| Technique \| Token Cost \| Reliability \| Use Case \|
69	\|------\|-----------\|------------\|-------------\|----------\|
70	\| Simple, well-defined task \| Zero-Shot \| ⭐⭐⭐⭐⭐ Minimal \| ⭐⭐⭐ Medium \| Translation, simple summarization \|
71	\| Specific format/style \| Few-Shot \| ⭐⭐⭐ Medium \| ⭐⭐⭐⭐ High \| Classification, entity extraction \|
72	\| Complex reasoning \| Chain-of-Thought \| ⭐⭐ Higher \| ⭐⭐⭐⭐⭐ Very High \| Math, logic, multi-hop QA \|
73	\| Structured data output \| JSON Mode / Tools \| ⭐⭐⭐⭐ Low-Med \| ⭐⭐⭐⭐⭐ Very High \| API responses, data extraction \|
74	\| Multi-step workflows \| Prompt Chaining \| ⭐⭐⭐ Medium \| ⭐⭐⭐⭐ High \| Pipelines, complex tasks \|
75	\| Knowledge retrieval \| RAG \| ⭐⭐ Higher \| ⭐⭐⭐⭐ High \| QA over documents \|
76	\| Agent behaviors \| ReAct (Tool Use) \| ⭐ Highest \| ⭐⭐⭐ Medium \| Multi-tool, complex tasks \|
77
78	Decision tree:
79	```
80	START
81	├─ Need structured JSON? → Use JSON Mode / Tool Calling (references/structured-outputs.md)
82	├─ Complex reasoning required? → Use Chain-of-Thought (references/chain-of-thought.md)
83	├─ Specific format/style needed? → Use Few-Shot Learning (references/few-shot-learning.md)
84	├─ Knowledge from documents? → Use RAG (references/rag-patterns.md)
85	├─ Multi-step workflow? → Use Prompt Chaining (references/prompt-chaining.md)
86	├─ Agent with tools? → Use Tool Use / ReAct (references/tool-use-guide.md)
87	└─ Simple task → Use Zero-Shot (references/zero-shot-patterns.md)
88	```
89
90	## Core Prompting Patterns
91
92	### 1. Zero-Shot Prompting
93
94	Pattern: Clear instruction + optional context + input + output format specification
95
96	When to use: Simple, well-defined tasks with clear expected outputs (summarization, translation, basic classification).
97
98	Best practices:
99	- Be specific about constraints and requirements
100	- Use imperative voice ("Summarize...", not "Can you summarize...")
101	- Specify output format upfront
102	- Set `temperature=0` for deterministic outputs
103
104	Example:
105	```python
106	prompt = """
107	Summarize the following customer review in 2 sentences, focusing on key concerns:
108
109	Review: [customer feedback text]
110
111	Summary:
112	"""
113	```
114
115	See `references/zero-shot-patterns.md` for comprehensive examples and anti-patterns.
116
117	##

People who liked this also viewed

6 picks

Type

prompt-engineering

62installs59forks

View on GitHub

How to install prompt-engineering?

ancoleman/ai-design-components/prompt-engineering

$npx -y skills add ancoleman/ai-design-components --skill prompt-engineering

Installs into the current project.

Summary

Files · 1

View on GitHub

SKILL.md

1	# Prompt Engineering
2
3	Design and optimize prompts for large language models (LLMs) to achieve reliable, high-quality outputs across diverse tasks.
4
5	## Purpose
6
7	This skill provides systematic techniques for crafting prompts that consistently elicit desired behaviors from LLMs. Rather than trial-and-error prompt iteration, apply proven patterns (zero-shot, few-shot, chain-of-thought, structured outputs) to improve accuracy, reduce costs, and build production-ready LLM applications. Covers multi-model deployment (OpenAI GPT, Anthropic Claude, Google Gemini, open-source models) with Python and TypeScript examples.
8
9	## When to Use This Skill
10
11	Trigger this skill when:
12	- Building LLM-powered applications requiring consistent outputs
13	- Model outputs are unreliable, inconsistent, or hallucinating
14	- Need structured data (JSON) from natural language inputs
15	- Implementing multi-step reasoning tasks (math, logic, analysis)
16	- Creating AI agents that use tools and external APIs
17	- Optimizing prompt costs or latency in production systems
18	- Migrating prompts across different model providers
19	- Establishing prompt versioning and testing workflows
20
21	Common requests:
22	- "How do I make Claude/GPT follow instructions reliably?"
23	- "My JSON parsing keeps failing - how to get valid outputs?"
24	- "Need to build a RAG system for question-answering"
25	- "How to reduce hallucination in model responses?"
26	- "What's the best way to implement multi-step workflows?"
27
28	## Quick Start
29
30	Zero-Shot Prompt (Python + OpenAI):
31	```python
32	from openai import OpenAI
33	client = OpenAI()
34
35	response = client.chat.completions.create(
36	model="gpt-4",
37	messages=[
38	{"role": "system", "content": "You are a helpful assistant."},
39	{"role": "user", "content": "Summarize this article in 3 sentences: [text]"}
40	],
41	temperature=0 # Deterministic output
42	)
43	print(response.choices[0].message.content)
44	```
45
46	Structured Output (TypeScript + Vercel AI SDK):
47	```typescript
48	import { generateObject } from 'ai';
49	import { openai } from '@ai-sdk/openai';
50	import { z } from 'zod';
51
52	const schema = z.object({
53	name: z.string(),
54	sentiment: z.enum(['positive', 'negative', 'neutral']),
55	});
56
57	const { object } = await generateObject({
58	model: openai('gpt-4'),
59	schema,
60	prompt: 'Extract sentiment from: "This product is amazing!"',
61	});
62	```
63
64	## Prompting Technique Decision Framework
65
66	Choose the right technique based on task requirements:
67
68	\| Goal \| Technique \| Token Cost \| Reliability \| Use Case \|
69	\|------\|-----------\|------------\|-------------\|----------\|
70	\| Simple, well-defined task \| Zero-Shot \| ⭐⭐⭐⭐⭐ Minimal \| ⭐⭐⭐ Medium \| Translation, simple summarization \|
71	\| Specific format/style \| Few-Shot \| ⭐⭐⭐ Medium \| ⭐⭐⭐⭐ High \| Classification, entity extraction \|
72	\| Complex reasoning \| Chain-of-Thought \| ⭐⭐ Higher \| ⭐⭐⭐⭐⭐ Very High \| Math, logic, multi-hop QA \|
73	\| Structured data output \| JSON Mode / Tools \| ⭐⭐⭐⭐ Low-Med \| ⭐⭐⭐⭐⭐ Very High \| API responses, data extraction \|
74	\| Multi-step workflows \| Prompt Chaining \| ⭐⭐⭐ Medium \| ⭐⭐⭐⭐ High \| Pipelines, complex tasks \|
75	\| Knowledge retrieval \| RAG \| ⭐⭐ Higher \| ⭐⭐⭐⭐ High \| QA over documents \|
76	\| Agent behaviors \| ReAct (Tool Use) \| ⭐ Highest \| ⭐⭐⭐ Medium \| Multi-tool, complex tasks \|
77
78	Decision tree:
79	```
80	START
81	├─ Need structured JSON? → Use JSON Mode / Tool Calling (references/structured-outputs.md)
82	├─ Complex reasoning required? → Use Chain-of-Thought (references/chain-of-thought.md)
83	├─ Specific format/style needed? → Use Few-Shot Learning (references/few-shot-learning.md)
84	├─ Knowledge from documents? → Use RAG (references/rag-patterns.md)
85	├─ Multi-step workflow? → Use Prompt Chaining (references/prompt-chaining.md)
86	├─ Agent with tools? → Use Tool Use / ReAct (references/tool-use-guide.md)
87	└─ Simple task → Use Zero-Shot (references/zero-shot-patterns.md)
88	```
89
90	## Core Prompting Patterns
91
92	### 1. Zero-Shot Prompting
93
94	Pattern: Clear instruction + optional context + input + output format specification
95
96	When to use: Simple, well-defined tasks with clear expected outputs (summarization, translation, basic classification).
97
98	Best practices:
99	- Be specific about constraints and requirements
100	- Use imperative voice ("Summarize...", not "Can you summarize...")
101	- Specify output format upfront
102	- Set `temperature=0` for deterministic outputs
103
104	Example:
105	```python
106	prompt = """
107	Summarize the following customer review in 2 sentences, focusing on key concerns:
108
109	Review: [customer feedback text]
110
111	Summary:
112	"""
113	```
114
115	See `references/zero-shot-patterns.md` for comprehensive examples and anti-patterns.
116
117	##

People who liked this also viewed

6 picks

Type