agent-tool-builder

Name: agent-tool-builder
Author: sickn33

44kstars6.5kforks

How to install agent-tool-builder?

sickn33/agentic-awesome-skills/agent-tool-builder

$npx -y skills add sickn33/agentic-awesome-skills --skill agent-tool-builder

Installs into the current project.

Summary

Tools are how AI agents interact with the world. A well-designed

Files · 1

View on GitHub

SKILL.md

1	# Agent Tool Builder
2
3	Tools are how AI agents interact with the world. A well-designed tool is the
4	difference between an agent that works and one that hallucinates, fails
5	silently, or costs 10x more tokens than necessary.
6
7	This skill covers tool design from schema to error handling. JSON Schema
8	best practices, description writing that actually helps the LLM, validation,
9	and the emerging MCP standard that's becoming the lingua franca for AI tools.
10
11	Key insight: Tool descriptions are more important than tool implementations.
12	The LLM never sees your code - it only sees the schema and description.
13
14	## Principles
15
16	- Description quality > implementation quality for LLM accuracy
17	- Aim for fewer than 20 tools - more causes confusion
18	- Every tool needs explicit error handling - silent failures poison agents
19	- Return strings, not objects - LLMs process text
20	- Validation gates before execution - reject, fix, or escalate, never silent fail
21	- Test tools with the LLM, not just unit tests
22
23	## Capabilities
24
25	- agent-tools
26	- function-calling
27	- tool-schema-design
28	- mcp-tools
29	- tool-validation
30	- tool-error-handling
31
32	## Scope
33
34	- multi-agent-coordination → multi-agent-orchestration
35	- agent-memory → agent-memory-systems
36	- api-design → api-designer
37	- llm-prompting → prompt-engineering
38
39	## Tooling
40
41	### Standards
42
43	- JSON Schema - When: All tool definitions Note: The universal format for tool schemas
44	- MCP (Model Context Protocol) - When: Building reusable, cross-platform tools Note: Anthropic's open standard, widely adopted
45
46	### Frameworks
47
48	- Anthropic SDK - When: Claude-based agents Note: Beta tool runner handles most complexity
49	- OpenAI Functions - When: OpenAI-based agents Note: Use strict mode for guaranteed schema compliance
50	- Vercel AI SDK - When: Multi-provider tool handling Note: Abstracts differences between providers
51	- LangChain Tools - When: LangChain-based agents Note: Converts MCP tools to LangChain format
52
53	## Patterns
54
55	### Tool Schema Design
56
57	Creating clear, unambiguous JSON Schema for tools
58
59	When to use: Defining any new tool for an agent
60
61	# TOOL SCHEMA BEST PRACTICES:
62
63	## 1. Detailed Descriptions (Most Important)
64	"""
65	BAD - Too vague:
66	{
67	"name": "get_stock_price",
68	"description": "Gets stock price",
69	"input_schema": {
70	"type": "object",
71	"properties": {
72	"ticker": {"type": "string"}
73	}
74	}
75	}
76
77	GOOD - Comprehensive:
78	{
79	"name": "get_stock_price",
80	"description": "Retrieves the current stock price for a given ticker
81	symbol. The ticker symbol must be a valid symbol for a publicly
82	traded company on a major US stock exchange like NYSE or NASDAQ.
83	Returns the latest trade price in USD. Use when the user asks
84	about current or recent stock prices. Does NOT provide historical
85	data, company info, or predictions.",
86	"input_schema": {
87	"type": "object",
88	"properties": {
89	"ticker": {
90	"type": "string",
91	"description": "The stock ticker symbol, e.g. AAPL for Apple Inc."
92	}
93	},
94	"required": ["ticker"]
95	}
96	}
97	"""
98
99	## 2. Parameter Descriptions
100	"""
101	Every parameter needs:
102	- What it is
103	- Format expected
104	- Example value
105	- Edge cases/limitations
106
107	{
108	"location": {
109	"type": "string",
110	"description": "City and state/country. Format: 'City, State' for US
111	(e.g., 'San Francisco, CA') or 'City, Country' for international
112	(e.g., 'Tokyo, Japan'). Do not use ZIP codes or coordinates."
113	},
114	"unit": {
115	"type": "string",
116	"enum": ["celsius", "fahrenheit"],
117	"description": "Temperature unit. Defaults to user's locale if not
118	specified. Use 'fahrenheit' for US users, 'celsius' for others."
119	}
120	}
121	"""
122
123	## 3. Use Enums When Possible
124	"""
125	Enums constrain the LLM to valid values:
126
127	"priority": {
128	"type": "string",
129	"enum": ["low", "medium", "high", "critical"],
130	"description": "Task priority level"
131	}
132
133	"action": {
134	"type": "string",
135	"enum": ["create", "read", "update", "delete"],
136	"description": "The CRUD operation to perform"
137	}
138	"""
139
140	## 4. Required vs Optional
141	"""
142	Be explicit about what's required:
143
144	{
145	"type": "object",
146	"properties": {
147	"query": {...}, // Required
148	"limit": {...}, // Optional with default
149	"offset": {...} // Optional
150	},
151	"required": ["query"],
152	"additionalProperties": false // Strict mode
153	}
154	"""
155
156	### Tool with Input Examples
157
158	Using examples to guide LLM tool usage
159
160	When to use: Complex tools with nested objects or format-sensitive inputs
161
162	# TOOL USE EXAMPLES (Anthropic Beta Feature):
163
164	"""
165	Examples show Claude concrete patterns that schemas can't express.
166	Improves accuracy from 72% to 90% on complex operatio

People who liked this also viewed

6 picks

Type

lark-openapi-explorer飞书/Lark 原生 OpenAPI 探索：从官方文档库中挖掘未经 CLI 封装的原生 OpenAPI 接口。当用户的需求无法被现有 lark-* skill 或 lark-cli 已注册命令满足，需要查找并调用原生飞书 OpenAPI 时使用。SkillsJul 2026378k16k
lark-skill-maker创建 lark-cli 的自定义 Skill。当用户需要把飞书 API 操作封装成可复用的 Skill（包装原子 API 或编排多步流程）时使用。SkillsJul 2026377k16k
azure-cost-optimizationOfficial agent plugin providing skills and MCP server configurations for Azure scenarios.SkillsJul 2026209k1.3k
vercel-react-native-skillsVercel's official collection of agent skillsSkillsJul 2026172k29k
mcp-builderGuide for creating high-quality MCP (Model Context Protocol) servers that enable LLMs to interact with external services through well-designed tools. Use when building MCP servers to integrate externaSkillsJul 202693k163k
gws-sharedgws CLI: Shared patterns for authentication, global flags, and output formatting.SkillsJul 202653k30k

agent-tool-builder

44kstars6.5kforks

View on GitHub

How to install agent-tool-builder?

sickn33/agentic-awesome-skills/agent-tool-builder

$npx -y skills add sickn33/agentic-awesome-skills --skill agent-tool-builder

Installs into the current project.

Summary

Tools are how AI agents interact with the world. A well-designed

Files · 1

View on GitHub

SKILL.md

1	# Agent Tool Builder
2
3	Tools are how AI agents interact with the world. A well-designed tool is the
4	difference between an agent that works and one that hallucinates, fails
5	silently, or costs 10x more tokens than necessary.
6
7	This skill covers tool design from schema to error handling. JSON Schema
8	best practices, description writing that actually helps the LLM, validation,
9	and the emerging MCP standard that's becoming the lingua franca for AI tools.
10
11	Key insight: Tool descriptions are more important than tool implementations.
12	The LLM never sees your code - it only sees the schema and description.
13
14	## Principles
15
16	- Description quality > implementation quality for LLM accuracy
17	- Aim for fewer than 20 tools - more causes confusion
18	- Every tool needs explicit error handling - silent failures poison agents
19	- Return strings, not objects - LLMs process text
20	- Validation gates before execution - reject, fix, or escalate, never silent fail
21	- Test tools with the LLM, not just unit tests
22
23	## Capabilities
24
25	- agent-tools
26	- function-calling
27	- tool-schema-design
28	- mcp-tools
29	- tool-validation
30	- tool-error-handling
31
32	## Scope
33
34	- multi-agent-coordination → multi-agent-orchestration
35	- agent-memory → agent-memory-systems
36	- api-design → api-designer
37	- llm-prompting → prompt-engineering
38
39	## Tooling
40
41	### Standards
42
43	- JSON Schema - When: All tool definitions Note: The universal format for tool schemas
44	- MCP (Model Context Protocol) - When: Building reusable, cross-platform tools Note: Anthropic's open standard, widely adopted
45
46	### Frameworks
47
48	- Anthropic SDK - When: Claude-based agents Note: Beta tool runner handles most complexity
49	- OpenAI Functions - When: OpenAI-based agents Note: Use strict mode for guaranteed schema compliance
50	- Vercel AI SDK - When: Multi-provider tool handling Note: Abstracts differences between providers
51	- LangChain Tools - When: LangChain-based agents Note: Converts MCP tools to LangChain format
52
53	## Patterns
54
55	### Tool Schema Design
56
57	Creating clear, unambiguous JSON Schema for tools
58
59	When to use: Defining any new tool for an agent
60
61	# TOOL SCHEMA BEST PRACTICES:
62
63	## 1. Detailed Descriptions (Most Important)
64	"""
65	BAD - Too vague:
66	{
67	"name": "get_stock_price",
68	"description": "Gets stock price",
69	"input_schema": {
70	"type": "object",
71	"properties": {
72	"ticker": {"type": "string"}
73	}
74	}
75	}
76
77	GOOD - Comprehensive:
78	{
79	"name": "get_stock_price",
80	"description": "Retrieves the current stock price for a given ticker
81	symbol. The ticker symbol must be a valid symbol for a publicly
82	traded company on a major US stock exchange like NYSE or NASDAQ.
83	Returns the latest trade price in USD. Use when the user asks
84	about current or recent stock prices. Does NOT provide historical
85	data, company info, or predictions.",
86	"input_schema": {
87	"type": "object",
88	"properties": {
89	"ticker": {
90	"type": "string",
91	"description": "The stock ticker symbol, e.g. AAPL for Apple Inc."
92	}
93	},
94	"required": ["ticker"]
95	}
96	}
97	"""
98
99	## 2. Parameter Descriptions
100	"""
101	Every parameter needs:
102	- What it is
103	- Format expected
104	- Example value
105	- Edge cases/limitations
106
107	{
108	"location": {
109	"type": "string",
110	"description": "City and state/country. Format: 'City, State' for US
111	(e.g., 'San Francisco, CA') or 'City, Country' for international
112	(e.g., 'Tokyo, Japan'). Do not use ZIP codes or coordinates."
113	},
114	"unit": {
115	"type": "string",
116	"enum": ["celsius", "fahrenheit"],
117	"description": "Temperature unit. Defaults to user's locale if not
118	specified. Use 'fahrenheit' for US users, 'celsius' for others."
119	}
120	}
121	"""
122
123	## 3. Use Enums When Possible
124	"""
125	Enums constrain the LLM to valid values:
126
127	"priority": {
128	"type": "string",
129	"enum": ["low", "medium", "high", "critical"],
130	"description": "Task priority level"
131	}
132
133	"action": {
134	"type": "string",
135	"enum": ["create", "read", "update", "delete"],
136	"description": "The CRUD operation to perform"
137	}
138	"""
139
140	## 4. Required vs Optional
141	"""
142	Be explicit about what's required:
143
144	{
145	"type": "object",
146	"properties": {
147	"query": {...}, // Required
148	"limit": {...}, // Optional with default
149	"offset": {...} // Optional
150	},
151	"required": ["query"],
152	"additionalProperties": false // Strict mode
153	}
154	"""
155
156	### Tool with Input Examples
157
158	Using examples to guide LLM tool usage
159
160	When to use: Complex tools with nested objects or format-sensitive inputs
161
162	# TOOL USE EXAMPLES (Anthropic Beta Feature):
163
164	"""
165	Examples show Claude concrete patterns that schemas can't express.
166	Improves accuracy from 72% to 90% on complex operatio

People who liked this also viewed

6 picks

Type