prompt-engineering

Name: prompt-engineering
Author: itsmostafa

25installs0forks

How to install prompt-engineering?

itsmostafa/llm-engineering-skills/prompt-engineering

$npx -y skills add itsmostafa/llm-engineering-skills --skill prompt-engineering

Installs into the current project.

Summary

Crafting effective prompts for LLMs. Use when designing prompts, improving output quality, structuring complex instructions, or debugging poor model responses.

Files · 1

View on GitHub

SKILL.md

1	# Prompt Engineering
2
3	Prompt engineering is the practice of designing inputs that guide LLMs to produce desired outputs. Effective prompts reduce errors, improve consistency, and unlock model capabilities.
4
5	## Table of Contents
6
7	- [Core Principles](#core-principles)
8	- [Be Clear and Direct](#be-clear-and-direct)
9	- [Use Examples (Multishot)](#use-examples-multishot)
10	- [Reasoning Guidance](#reasoning-guidance)
11	- [XML Tags](#xml-tags)
12	- [Role Prompting](#role-prompting)
13	- [Long Context](#long-context)
14	- [Output Control](#output-control)
15	- [Self-Verification](#self-verification)
16	- [Best Practices](#best-practices)
17	- [References](#references)
18
19	## Core Principles
20
21	Golden rule: Show your prompt to a colleague with minimal context. If they're confused, the model will be too.
22
23	1. Be explicit - State exactly what you want; never assume the model knows your preferences
24	2. Provide context - Include what the output is for, who the audience is, and what success looks like
25	3. Use structure - Sequential steps, XML tags, and clear formatting reduce ambiguity
26	4. Show examples - Demonstrations outperform descriptions for complex formats
27
28	## Be Clear and Direct
29
30	Treat the model as a capable but context-free collaborator. Specify:
31	- What the task results will be used for
32	- What audience the output is meant for
33	- What a successful completion looks like
34
35	### Vague vs Specific
36
37	```
38	# Vague
39	Analyze this data and give insights.
40
41	# Specific
42	Analyze this Q2 sales data for our board presentation.
43	1. Identify the top 3 revenue trends
44	2. Flag any anomalies exceeding 15% variance
45	3. Recommend 2-3 actionable next steps
46	Format as bullet points, max 200 words.
47	```
48
49	### Sequential Steps
50
51	Use numbered lists for multi-step tasks:
52
53	```
54	Your task is to anonymize customer feedback.
55
56	Instructions:
57	1. Replace customer names with "CUSTOMER_[ID]"
58	2. Replace emails with "EMAIL_[ID]@example.com"
59	3. Redact phone numbers as "PHONE_[ID]"
60	4. Leave product names intact
61	5. Output only processed messages, separated by "---"
62	```
63
64	## Use Examples (Multishot)
65
66	Provide 3-5 diverse examples to demonstrate expected behavior. Examples reduce misinterpretation and enforce consistent formatting.
67
68	### Structure
69
70	```
71	Categorize customer feedback by issue type and sentiment.
72
73	<examples>
74	<example>
75	Input: The dashboard loads slowly and the export button is hidden.
76	Category: UI/UX, Performance
77	Sentiment: Negative
78	Priority: High
79	</example>
80
81	<example>
82	Input: Love the Salesforce integration! Would be great to add Hubspot.
83	Category: Integration, Feature Request
84	Sentiment: Positive
85	Priority: Medium
86	</example>
87	</examples>
88
89	Now categorize: {{FEEDBACK}}
90	```
91
92	### Tips
93
94	- Make examples relevant to actual use cases
95	- Include edge cases and potential challenges
96	- Vary examples to prevent unintended pattern matching
97	- Wrap in `<example>` tags for clarity
98
99	## Reasoning Guidance
100
101	For complex tasks, guide the model to reason before answering, but ask for concise conclusions or verifiable work rather than unrestricted hidden chain-of-thought.
102
103	### Basic
104
105	```
106	Determine the best investment option for this client. Think step-by-step.
107	```
108
109	### Guided
110
111	Specify what steps to consider:
112
113	```
114	Think before answering:
115	1. Consider the client's risk tolerance given their 5-year timeline
116	2. Calculate potential returns for each option
117	3. Factor in market volatility history
118	4. Then provide your recommendation
119	```
120
121	### Structured (Recommended)
122
123	Separate private analysis instructions from the required answer format:
124
125	```
126	Analyze this contract for legal risks.
127
128	Before answering, check:
129	- Indemnification implications
130	- Liability exposure
131	- IP ownership concerns
132
133	Then provide:
134	<answer>
135	- Top risks
136	- Recommended edits
137	- Open questions
138	</answer>
139	```
140
141	When you need auditability, ask for brief supporting rationale, calculations, cited evidence, or a checklist of checks performed. Do not require full chain-of-thought unless the target model or product explicitly supports exposing it.
142
143	## XML Tags
144
145	Use XML tags to separate prompt components. This prevents instruction/content confusion and improves parseability.
146
147	### Common Tags
148
149	```
150	<instructions>Task steps and requirements</instructions>
151	<context>Background information</context>
152	<document>Source material to process</document>
153	<example>Demonstration of expected behavior</example>
154	<constraints>Boundaries and limitations</constraints>
155	<output_format>Expected response structure</output_format>
156	```
157
158	### Nested Structure
159
160	```
161	<documents>
162	<document index="1">
163	<source>annual_report_2023.pdf</source>
164	<content>{{REPORT_CONTENT}}</content>
165	</document>
166	<document index="2">
167	<source>competitor_analysis.xlsx</source>
168	<content>{{ANALYSIS_CONTENT}}</content>
169	</document>
170	</documents>
171
172	<instructions>
173	Compare revenue trends across bo

People who liked this also viewed

6 picks

Type

prompt-engineering

25installs0forks

View on GitHub

How to install prompt-engineering?

itsmostafa/llm-engineering-skills/prompt-engineering

$npx -y skills add itsmostafa/llm-engineering-skills --skill prompt-engineering

Installs into the current project.

Summary

Crafting effective prompts for LLMs. Use when designing prompts, improving output quality, structuring complex instructions, or debugging poor model responses.

Files · 1

View on GitHub

SKILL.md

1	# Prompt Engineering
2
3	Prompt engineering is the practice of designing inputs that guide LLMs to produce desired outputs. Effective prompts reduce errors, improve consistency, and unlock model capabilities.
4
5	## Table of Contents
6
7	- [Core Principles](#core-principles)
8	- [Be Clear and Direct](#be-clear-and-direct)
9	- [Use Examples (Multishot)](#use-examples-multishot)
10	- [Reasoning Guidance](#reasoning-guidance)
11	- [XML Tags](#xml-tags)
12	- [Role Prompting](#role-prompting)
13	- [Long Context](#long-context)
14	- [Output Control](#output-control)
15	- [Self-Verification](#self-verification)
16	- [Best Practices](#best-practices)
17	- [References](#references)
18
19	## Core Principles
20
21	Golden rule: Show your prompt to a colleague with minimal context. If they're confused, the model will be too.
22
23	1. Be explicit - State exactly what you want; never assume the model knows your preferences
24	2. Provide context - Include what the output is for, who the audience is, and what success looks like
25	3. Use structure - Sequential steps, XML tags, and clear formatting reduce ambiguity
26	4. Show examples - Demonstrations outperform descriptions for complex formats
27
28	## Be Clear and Direct
29
30	Treat the model as a capable but context-free collaborator. Specify:
31	- What the task results will be used for
32	- What audience the output is meant for
33	- What a successful completion looks like
34
35	### Vague vs Specific
36
37	```
38	# Vague
39	Analyze this data and give insights.
40
41	# Specific
42	Analyze this Q2 sales data for our board presentation.
43	1. Identify the top 3 revenue trends
44	2. Flag any anomalies exceeding 15% variance
45	3. Recommend 2-3 actionable next steps
46	Format as bullet points, max 200 words.
47	```
48
49	### Sequential Steps
50
51	Use numbered lists for multi-step tasks:
52
53	```
54	Your task is to anonymize customer feedback.
55
56	Instructions:
57	1. Replace customer names with "CUSTOMER_[ID]"
58	2. Replace emails with "EMAIL_[ID]@example.com"
59	3. Redact phone numbers as "PHONE_[ID]"
60	4. Leave product names intact
61	5. Output only processed messages, separated by "---"
62	```
63
64	## Use Examples (Multishot)
65
66	Provide 3-5 diverse examples to demonstrate expected behavior. Examples reduce misinterpretation and enforce consistent formatting.
67
68	### Structure
69
70	```
71	Categorize customer feedback by issue type and sentiment.
72
73	<examples>
74	<example>
75	Input: The dashboard loads slowly and the export button is hidden.
76	Category: UI/UX, Performance
77	Sentiment: Negative
78	Priority: High
79	</example>
80
81	<example>
82	Input: Love the Salesforce integration! Would be great to add Hubspot.
83	Category: Integration, Feature Request
84	Sentiment: Positive
85	Priority: Medium
86	</example>
87	</examples>
88
89	Now categorize: {{FEEDBACK}}
90	```
91
92	### Tips
93
94	- Make examples relevant to actual use cases
95	- Include edge cases and potential challenges
96	- Vary examples to prevent unintended pattern matching
97	- Wrap in `<example>` tags for clarity
98
99	## Reasoning Guidance
100
101	For complex tasks, guide the model to reason before answering, but ask for concise conclusions or verifiable work rather than unrestricted hidden chain-of-thought.
102
103	### Basic
104
105	```
106	Determine the best investment option for this client. Think step-by-step.
107	```
108
109	### Guided
110
111	Specify what steps to consider:
112
113	```
114	Think before answering:
115	1. Consider the client's risk tolerance given their 5-year timeline
116	2. Calculate potential returns for each option
117	3. Factor in market volatility history
118	4. Then provide your recommendation
119	```
120
121	### Structured (Recommended)
122
123	Separate private analysis instructions from the required answer format:
124
125	```
126	Analyze this contract for legal risks.
127
128	Before answering, check:
129	- Indemnification implications
130	- Liability exposure
131	- IP ownership concerns
132
133	Then provide:
134	<answer>
135	- Top risks
136	- Recommended edits
137	- Open questions
138	</answer>
139	```
140
141	When you need auditability, ask for brief supporting rationale, calculations, cited evidence, or a checklist of checks performed. Do not require full chain-of-thought unless the target model or product explicitly supports exposing it.
142
143	## XML Tags
144
145	Use XML tags to separate prompt components. This prevents instruction/content confusion and improves parseability.
146
147	### Common Tags
148
149	```
150	<instructions>Task steps and requirements</instructions>
151	<context>Background information</context>
152	<document>Source material to process</document>
153	<example>Demonstration of expected behavior</example>
154	<constraints>Boundaries and limitations</constraints>
155	<output_format>Expected response structure</output_format>
156	```
157
158	### Nested Structure
159
160	```
161	<documents>
162	<document index="1">
163	<source>annual_report_2023.pdf</source>
164	<content>{{REPORT_CONTENT}}</content>
165	</document>
166	<document index="2">
167	<source>competitor_analysis.xlsx</source>
168	<content>{{ANALYSIS_CONTENT}}</content>
169	</document>
170	</documents>
171
172	<instructions>
173	Compare revenue trends across bo

People who liked this also viewed

6 picks

Type