autonomous-agent-patterns

Name: autonomous-agent-patterns
Author: SamarthaKV29

64stars12forks

How to install autonomous-agent-patterns?

samarthakv29/antigravity-god-mode/autonomous-agent-patterns

$npx -y skills add SamarthaKV29/antigravity-god-mode --skill autonomous-agent-patterns

Installs into the current project.

Summary

Design patterns for building autonomous coding agents. Covers tool integration, permission systems, browser automation, and human-in-the-loop workflows. Use when building AI agents, designing tool APIs, implementing permission systems, or creating autonomous coding assistants.

Files · 1

View on GitHub

SKILL.md

1	# 🕹️ Autonomous Agent Patterns
2
3	> Design patterns for building autonomous coding agents, inspired by [Cline](https://github.com/cline/cline) and [OpenAI Codex](https://github.com/openai/codex).
4
5	## When to Use This Skill
6
7	Use this skill when:
8
9	- Building autonomous AI agents
10	- Designing tool/function calling APIs
11	- Implementing permission and approval systems
12	- Creating browser automation for agents
13	- Designing human-in-the-loop workflows
14
15	---
16
17	## 1. Core Agent Architecture
18
19	### 1.1 Agent Loop
20
21	```
22	┌─────────────────────────────────────────────────────────────┐
23	│ AGENT LOOP │
24	│ │
25	│ ┌──────────┐ ┌──────────┐ ┌──────────┐ │
26	│ │ Think │───▶│ Decide │───▶│ Act │ │
27	│ │ (Reason) │ │ (Plan) │ │ (Execute)│ │
28	│ └──────────┘ └──────────┘ └──────────┘ │
29	│ ▲ │ │
30	│ │ ┌──────────┐ │ │
31	│ └─────────│ Observe │◀─────────┘ │
32	│ │ (Result) │ │
33	│ └──────────┘ │
34	└─────────────────────────────────────────────────────────────┘
35	```
36
37	```python
38	class AgentLoop:
39	def __init__(self, llm, tools, max_iterations=50):
40	self.llm = llm
41	self.tools = {t.name: t for t in tools}
42	self.max_iterations = max_iterations
43	self.history = []
44
45	def run(self, task: str) -> str:
46	self.history.append({"role": "user", "content": task})
47
48	for i in range(self.max_iterations):
49	# Think: Get LLM response with tool options
50	response = self.llm.chat(
51	messages=self.history,
52	tools=self._format_tools(),
53	tool_choice="auto"
54	)
55
56	# Decide: Check if agent wants to use a tool
57	if response.tool_calls:
58	for tool_call in response.tool_calls:
59	# Act: Execute the tool
60	result = self._execute_tool(tool_call)
61
62	# Observe: Add result to history
63	self.history.append({
64	"role": "tool",
65	"tool_call_id": tool_call.id,
66	"content": str(result)
67	})
68	else:
69	# No more tool calls = task complete
70	return response.content
71
72	return "Max iterations reached"
73
74	def _execute_tool(self, tool_call) -> Any:
75	tool = self.tools[tool_call.name]
76	args = json.loads(tool_call.arguments)
77	return tool.execute(**args)
78	```
79
80	### 1.2 Multi-Model Architecture
81
82	```python
83	class MultiModelAgent:
84	"""
85	Use different models for different purposes:
86	- Fast model for planning
87	- Powerful model for complex reasoning
88	- Specialized model for code generation
89	"""
90
91	def __init__(self):
92	self.models = {
93	"fast": "gpt-3.5-turbo", # Quick decisions
94	"smart": "gpt-4-turbo", # Complex reasoning
95	"code": "claude-3-sonnet", # Code generation
96	}
97
98	def select_model(self, task_type: str) -> str:
99	if task_type == "planning":
100	return self.models["fast"]
101	elif task_type == "analysis":
102	return self.models["smart"]
103	elif task_type == "code":
104	return self.models["code"]
105	return self.models["smart"]
106	```
107
108	---
109
110	## 2. Tool Design Patterns
111
112	### 2.1 Tool Schema
113
114	```python
115	class Tool:
116	"""Base class for agent tools"""
117
118	@property
119	def schema(self) -> dict:
120	"""JSON Schema for the tool"""
121	return {
122	"name": self.name,
123	"description": self.description,
124	"parameters": {
125	"type": "object",
126	"properties": self._get_parameters(),
127	"required": self._get_required()
128	}
129	}
130
131	def execute(self, **kwargs) -> ToolResult:
132	"""Execute the tool and return result"""
133	raise NotImplementedError
134
135	class ReadFileTool(Tool):
136	name = "read_file"
137	description = "Read the contents of a file from the filesystem"
138
139	def _get_parameters(self):
140	return {
141	"path": {
142	"type": "string",
143	"description": "Absolute path to the file"
144	},
145	"start_line": {
146	"type": "integer",
147	"description": "Line to start reading from (1-indexed)"
148	},
149	"end_lin

People who liked this also viewed

6 picks

Type

autonomous-agent-patterns

64stars12forks

View on GitHub

How to install autonomous-agent-patterns?

samarthakv29/antigravity-god-mode/autonomous-agent-patterns

$npx -y skills add SamarthaKV29/antigravity-god-mode --skill autonomous-agent-patterns

Installs into the current project.

Summary

Files · 1

View on GitHub

SKILL.md

1	# 🕹️ Autonomous Agent Patterns
2
3	> Design patterns for building autonomous coding agents, inspired by [Cline](https://github.com/cline/cline) and [OpenAI Codex](https://github.com/openai/codex).
4
5	## When to Use This Skill
6
7	Use this skill when:
8
9	- Building autonomous AI agents
10	- Designing tool/function calling APIs
11	- Implementing permission and approval systems
12	- Creating browser automation for agents
13	- Designing human-in-the-loop workflows
14
15	---
16
17	## 1. Core Agent Architecture
18
19	### 1.1 Agent Loop
20
21	```
22	┌─────────────────────────────────────────────────────────────┐
23	│ AGENT LOOP │
24	│ │
25	│ ┌──────────┐ ┌──────────┐ ┌──────────┐ │
26	│ │ Think │───▶│ Decide │───▶│ Act │ │
27	│ │ (Reason) │ │ (Plan) │ │ (Execute)│ │
28	│ └──────────┘ └──────────┘ └──────────┘ │
29	│ ▲ │ │
30	│ │ ┌──────────┐ │ │
31	│ └─────────│ Observe │◀─────────┘ │
32	│ │ (Result) │ │
33	│ └──────────┘ │
34	└─────────────────────────────────────────────────────────────┘
35	```
36
37	```python
38	class AgentLoop:
39	def __init__(self, llm, tools, max_iterations=50):
40	self.llm = llm
41	self.tools = {t.name: t for t in tools}
42	self.max_iterations = max_iterations
43	self.history = []
44
45	def run(self, task: str) -> str:
46	self.history.append({"role": "user", "content": task})
47
48	for i in range(self.max_iterations):
49	# Think: Get LLM response with tool options
50	response = self.llm.chat(
51	messages=self.history,
52	tools=self._format_tools(),
53	tool_choice="auto"
54	)
55
56	# Decide: Check if agent wants to use a tool
57	if response.tool_calls:
58	for tool_call in response.tool_calls:
59	# Act: Execute the tool
60	result = self._execute_tool(tool_call)
61
62	# Observe: Add result to history
63	self.history.append({
64	"role": "tool",
65	"tool_call_id": tool_call.id,
66	"content": str(result)
67	})
68	else:
69	# No more tool calls = task complete
70	return response.content
71
72	return "Max iterations reached"
73
74	def _execute_tool(self, tool_call) -> Any:
75	tool = self.tools[tool_call.name]
76	args = json.loads(tool_call.arguments)
77	return tool.execute(**args)
78	```
79
80	### 1.2 Multi-Model Architecture
81
82	```python
83	class MultiModelAgent:
84	"""
85	Use different models for different purposes:
86	- Fast model for planning
87	- Powerful model for complex reasoning
88	- Specialized model for code generation
89	"""
90
91	def __init__(self):
92	self.models = {
93	"fast": "gpt-3.5-turbo", # Quick decisions
94	"smart": "gpt-4-turbo", # Complex reasoning
95	"code": "claude-3-sonnet", # Code generation
96	}
97
98	def select_model(self, task_type: str) -> str:
99	if task_type == "planning":
100	return self.models["fast"]
101	elif task_type == "analysis":
102	return self.models["smart"]
103	elif task_type == "code":
104	return self.models["code"]
105	return self.models["smart"]
106	```
107
108	---
109
110	## 2. Tool Design Patterns
111
112	### 2.1 Tool Schema
113
114	```python
115	class Tool:
116	"""Base class for agent tools"""
117
118	@property
119	def schema(self) -> dict:
120	"""JSON Schema for the tool"""
121	return {
122	"name": self.name,
123	"description": self.description,
124	"parameters": {
125	"type": "object",
126	"properties": self._get_parameters(),
127	"required": self._get_required()
128	}
129	}
130
131	def execute(self, **kwargs) -> ToolResult:
132	"""Execute the tool and return result"""
133	raise NotImplementedError
134
135	class ReadFileTool(Tool):
136	name = "read_file"
137	description = "Read the contents of a file from the filesystem"
138
139	def _get_parameters(self):
140	return {
141	"path": {
142	"type": "string",
143	"description": "Absolute path to the file"
144	},
145	"start_line": {
146	"type": "integer",
147	"description": "Line to start reading from (1-indexed)"
148	},
149	"end_lin

People who liked this also viewed

6 picks

Type