create-agent

Name: create-agent
Author: NeoLabHQ

829installs132forks

How to install create-agent?

neolabhq/context-engineering-kit/create-agent

$npx -y skills add NeoLabHQ/context-engineering-kit --skill create-agent

Installs into the current project.

Summary

Comprehensive guide for creating Claude Code agents with proper structure, triggering conditions, system prompts, and validation - combines official Anthropic best practices with proven patterns

Files · 1

View on GitHub

SKILL.md

1	# Create Agent Command
2
3	Create autonomous Claude Code agents that handle complex, multi-step tasks independently. This command provides comprehensive guidance based on official Anthropic documentation and proven patterns.
4
5	## User Input
6
7	```text
8	Agent Name: $1
9	Description: $2
10	```
11
12	## What Are Agents?
13
14	Agents are autonomous subprocesses spawned via the Task tool that:
15
16	- Handle complex, multi-step tasks independently
17	- Have their own isolated context window
18	- Return results to the parent conversation
19	- Can be specialized for specific domains
20
21	\| Concept \| Agent \| Command \|
22	\|---------\|-------\|---------\|
23	\| Trigger \| Claude decides based on description \| User invokes with `/name` \|
24	\| Purpose \| Autonomous work \| User-initiated actions \|
25	\| Context \| Isolated subprocess \| Shared conversation \|
26	\| File format \| `agents/.md` \| `commands/.md` \|
27
28	## Agent File Structure
29
30	Agents use a unique format combining YAML frontmatter with a markdown system prompt:
31
32	```markdown
33	---
34	name: agent-identifier
35	description: Use this agent when [triggering conditions]. Examples:
36
37	<example>
38	Context: [Situation description]
39	user: "[User request]"
40	assistant: "[How assistant should respond and use this agent]"
41	<commentary>
42	[Why this agent should be triggered]
43	</commentary>
44	</example>
45
46	<example>
47	[Additional example...]
48	</example>
49
50	model: inherit
51	color: blue
52	tools: ["Read", "Write", "Grep"]
53	---
54
55	You are [agent role description]...
56
57	Your Core Responsibilities:
58	1. [Responsibility 1]
59	2. [Responsibility 2]
60
61	Analysis Process:
62	[Step-by-step workflow]
63
64	Output Format:
65	[What to return]
66	```
67
68	## Frontmatter Fields Reference
69
70	### Required Fields
71
72	#### `name` (Required)
73
74	Format: Lowercase with hyphens only
75	Length: 3-50 characters
76	Rules:
77
78	- Must start and end with alphanumeric character
79	- Only lowercase letters, numbers, and hyphens
80	- No underscores, spaces, or special characters
81
82	\| Valid \| Invalid \| Reason \|
83	\|-------\|---------\|--------\|
84	\| `code-reviewer` \| `helper` \| Too generic \|
85	\| `test-generator` \| `-agent-` \| Starts/ends with hyphen \|
86	\| `api-docs-writer` \| `my_agent` \| Underscores not allowed \|
87	\| `security-analyzer` \| `ag` \| Too short (<3 chars) \|
88	\| `pr-quality-reviewer` \| `MyAgent` \| Uppercase not allowed \|
89
90	#### `description` (Required, Critical)
91
92	The most important field - Defines when Claude triggers the agent.
93
94	Requirements:
95
96	- Length: 10-5,000 characters (ideal: 200-1,000 with 2-4 examples)
97	- MUST start with: "Use this agent when..."
98	- MUST include: `<example>` blocks showing usage patterns
99	- Each example needs: context, user request, assistant response, commentary
100
101	Example Block Format:
102
103	```markdown
104	<example>
105	Context: [Describe the situation - what led to this interaction]
106	user: "[Exact user message or request]"
107	assistant: "[How Claude should respond before triggering]"
108	<commentary>
109	[Explanation of why this agent should be triggered in this scenario]
110	</commentary>
111	assistant: "[How Claude triggers the agent - 'I'll use the [agent-name] agent...']"
112	</example>
113	```
114
115	Best Practices for Descriptions:
116
117	- Include 2-4 concrete examples
118	- Show both proactive and reactive triggering scenarios
119	- Cover different phrasings of the same intent
120	- Explain reasoning in commentary
121	- Be specific about when NOT to use the agent
122
123	#### `model` (Required)
124
125	Values: `inherit`, `sonnet`, `opus`, `haiku`
126	Default: `inherit` (recommended)
127
128	\| Value \| Use Case \| Cost \|
129	\|-------\|----------\|------\|
130	\| `inherit` \| Use parent conversation model \| Default \|
131	\| `haiku` \| Fast, simple tasks \| Lowest \|
132	\| `sonnet` \| Balanced performance \| Medium \|
133	\| `opus` \| Maximum capability, complex reasoning \| Highest \|
134
135	Recommendation: Use `inherit` unless you have a specific reason to override.
136
137	#### `color` (Required)
138
139	Purpose: Visual indicator in UI to distinguish agents
140
141	Values: `blue`, `cyan`, `green`, `yellow`, `magenta`, `red`
142
143	Best Practice: Use different colors for different agents in the same plugin to distinguish them visually.
144
145	### Optional Fields
146
147	#### `tools` (Optional)
148
149	Purpose: Restrict available tools (principle of least privilege)
150	Format: Array of tool names
151	Default: All tools available
152
153	```yaml
154	# Common tool restrictions
155	tools: ["Read", "Grep", "Glob"] # Read-only analysis
156	tools: ["Read", "Write", "Grep", "Glob"] # Code modification
157	tools: ["Read", "Bash", "Grep"] # System operations
158	```
159
160	Security Principle: Only grant tools the agent actually needs.
161
162	## Triggering Patterns
163
164	### Pattern 1: Explicit Request
165
166	User directly asks for the agent's function.
167
168	```markdown
169	<example>
170	Context: User

People who liked this also viewed

6 picks

Type

create-agent

829installs132forks

View on GitHub

How to install create-agent?

neolabhq/context-engineering-kit/create-agent

$npx -y skills add NeoLabHQ/context-engineering-kit --skill create-agent

Installs into the current project.

Summary

Comprehensive guide for creating Claude Code agents with proper structure, triggering conditions, system prompts, and validation - combines official Anthropic best practices with proven patterns

Files · 1

View on GitHub

SKILL.md

1	# Create Agent Command
2
3	Create autonomous Claude Code agents that handle complex, multi-step tasks independently. This command provides comprehensive guidance based on official Anthropic documentation and proven patterns.
4
5	## User Input
6
7	```text
8	Agent Name: $1
9	Description: $2
10	```
11
12	## What Are Agents?
13
14	Agents are autonomous subprocesses spawned via the Task tool that:
15
16	- Handle complex, multi-step tasks independently
17	- Have their own isolated context window
18	- Return results to the parent conversation
19	- Can be specialized for specific domains
20
21	\| Concept \| Agent \| Command \|
22	\|---------\|-------\|---------\|
23	\| Trigger \| Claude decides based on description \| User invokes with `/name` \|
24	\| Purpose \| Autonomous work \| User-initiated actions \|
25	\| Context \| Isolated subprocess \| Shared conversation \|
26	\| File format \| `agents/.md` \| `commands/.md` \|
27
28	## Agent File Structure
29
30	Agents use a unique format combining YAML frontmatter with a markdown system prompt:
31
32	```markdown
33	---
34	name: agent-identifier
35	description: Use this agent when [triggering conditions]. Examples:
36
37	<example>
38	Context: [Situation description]
39	user: "[User request]"
40	assistant: "[How assistant should respond and use this agent]"
41	<commentary>
42	[Why this agent should be triggered]
43	</commentary>
44	</example>
45
46	<example>
47	[Additional example...]
48	</example>
49
50	model: inherit
51	color: blue
52	tools: ["Read", "Write", "Grep"]
53	---
54
55	You are [agent role description]...
56
57	Your Core Responsibilities:
58	1. [Responsibility 1]
59	2. [Responsibility 2]
60
61	Analysis Process:
62	[Step-by-step workflow]
63
64	Output Format:
65	[What to return]
66	```
67
68	## Frontmatter Fields Reference
69
70	### Required Fields
71
72	#### `name` (Required)
73
74	Format: Lowercase with hyphens only
75	Length: 3-50 characters
76	Rules:
77
78	- Must start and end with alphanumeric character
79	- Only lowercase letters, numbers, and hyphens
80	- No underscores, spaces, or special characters
81
82	\| Valid \| Invalid \| Reason \|
83	\|-------\|---------\|--------\|
84	\| `code-reviewer` \| `helper` \| Too generic \|
85	\| `test-generator` \| `-agent-` \| Starts/ends with hyphen \|
86	\| `api-docs-writer` \| `my_agent` \| Underscores not allowed \|
87	\| `security-analyzer` \| `ag` \| Too short (<3 chars) \|
88	\| `pr-quality-reviewer` \| `MyAgent` \| Uppercase not allowed \|
89
90	#### `description` (Required, Critical)
91
92	The most important field - Defines when Claude triggers the agent.
93
94	Requirements:
95
96	- Length: 10-5,000 characters (ideal: 200-1,000 with 2-4 examples)
97	- MUST start with: "Use this agent when..."
98	- MUST include: `<example>` blocks showing usage patterns
99	- Each example needs: context, user request, assistant response, commentary
100
101	Example Block Format:
102
103	```markdown
104	<example>
105	Context: [Describe the situation - what led to this interaction]
106	user: "[Exact user message or request]"
107	assistant: "[How Claude should respond before triggering]"
108	<commentary>
109	[Explanation of why this agent should be triggered in this scenario]
110	</commentary>
111	assistant: "[How Claude triggers the agent - 'I'll use the [agent-name] agent...']"
112	</example>
113	```
114
115	Best Practices for Descriptions:
116
117	- Include 2-4 concrete examples
118	- Show both proactive and reactive triggering scenarios
119	- Cover different phrasings of the same intent
120	- Explain reasoning in commentary
121	- Be specific about when NOT to use the agent
122
123	#### `model` (Required)
124
125	Values: `inherit`, `sonnet`, `opus`, `haiku`
126	Default: `inherit` (recommended)
127
128	\| Value \| Use Case \| Cost \|
129	\|-------\|----------\|------\|
130	\| `inherit` \| Use parent conversation model \| Default \|
131	\| `haiku` \| Fast, simple tasks \| Lowest \|
132	\| `sonnet` \| Balanced performance \| Medium \|
133	\| `opus` \| Maximum capability, complex reasoning \| Highest \|
134
135	Recommendation: Use `inherit` unless you have a specific reason to override.
136
137	#### `color` (Required)
138
139	Purpose: Visual indicator in UI to distinguish agents
140
141	Values: `blue`, `cyan`, `green`, `yellow`, `magenta`, `red`
142
143	Best Practice: Use different colors for different agents in the same plugin to distinguish them visually.
144
145	### Optional Fields
146
147	#### `tools` (Optional)
148
149	Purpose: Restrict available tools (principle of least privilege)
150	Format: Array of tool names
151	Default: All tools available
152
153	```yaml
154	# Common tool restrictions
155	tools: ["Read", "Grep", "Glob"] # Read-only analysis
156	tools: ["Read", "Write", "Grep", "Glob"] # Code modification
157	tools: ["Read", "Bash", "Grep"] # System operations
158	```
159
160	Security Principle: Only grant tools the agent actually needs.
161
162	## Triggering Patterns
163
164	### Pattern 1: Explicit Request
165
166	User directly asks for the agent's function.
167
168	```markdown
169	<example>
170	Context: User

People who liked this also viewed

6 picks

Type