Agent Browser
You are an expert python engineer. Automates browser interactions for web testing, form.
Before Starting
- Goal — what specific outcome do you need?
- Environment — versions, platform, existing setup?
- Constraints — performance, security, compatibility requirements?
- Integration — what systems does this connect to?
- Output format — code, config, script, or documentation?
Core Expertise Areas
- Core implementation — full working code for Agent Browser
- Error handling — robust error recovery and logging
- Performance — optimized patterns for production use
- Testing — unit and integration test strategies
- Configuration — environment-specific setup and tuning
- Security — secure coding patterns and best practices
- Documentation — clear API and usage documentation
Key Patterns & Code
Core Implementation
import anthropic
import json
client = anthropic.Anthropic()
TOOLS = [
{
"name": "search",
"description": "Search for information on a topic",
"input_schema": {
"type": "object",
"properties": {
"query": {"type": "string", "description": "Search query"}
},
"required": ["query"]
}
}
]
def handle_tool(name: str, inputs: dict) -> str:
if name == "search":
return f"Search results for: {inputs['query']}"
return f"Unknown tool: {name}"
def run_agent_browser(task: str, max_turns: int = 10) -> str:
messages = [{"role": "user", "content": task}]
for _ in range(max_turns):
response = client.messages.create(
model="claude-sonnet-4-6",
max_tokens=4096,
tools=TOOLS,
messages=messages
)
messages.append({"role": "assistant", "content": response.content})
if response.stop_reason == "end_turn":
return next((b.text for b in response.content if hasattr(b, 'text')), "")
if response.stop_reason == "tool_use":
results = [
{"type": "tool_result", "tool_use_id": b.id,
"content": handle_tool(b.name, b.input)}
for b in response.content if b.type == "tool_use"
]
messages.append({"role": "user", "content": results})
return "Max turns reached"
if __name__ == "__main__":
result = run_agent_browser("AgentBrowser task: analyze and report")
print(result)
Configuration & Setup
# Agent Browser — Configuration
# Author: luo-kai (Lous Creations)
config = {
"name": "agent-browser",
"version": "1.0.0",
"author": "luo-kai",
"enabled": True,
"debug": False,
"timeout_seconds": 30,
"max_retries": 3,
}
Error Handling
# Robust error handling pattern
import logging
logger = logging.getLogger("agent-browser")
def safe_run(func, *args, **kwargs):
try:
return func(*args, **kwargs)
except Exception as e:
logger.error(f"agent-browser error: {e}", exc_info=True)
raise
Best Practices
- Fail fast with clear errors — raise descriptive exceptions with context
- Log at appropriate levels — DEBUG for dev, INFO for ops, ERROR for problems
- Validate inputs — never trust external data without validation
- Use type annotations — improves IDE support and catches bugs early
- Handle cleanup — use context managers and
finally blocks
- Test edge cases — empty inputs, nulls, max values, concurrent access
Common Pitfalls
| Pitfall |
Problem |
Fix |
| No error handling |
Silent failures in production |
Wrap with try/except + logging |
| Hardcoded values |
Not portable across environments |
Use config/env vars |
| Missing timeouts |
Hangs indefinitely |
Always set timeout values |
| No retry logic |
Single failure = broken workflow |
Add exponential backoff |
| No cleanup on exit |
Resource leaks |
Use context managers |
Related Skills
- python-expert
- agent-browser-advanced
- performance-optimization
- error-handling
- testing-expert
1---2author: luo-kai3name: oc-agent-browser-24version: 1.0.05description: Automates browser interactions for web testing, form.6license: MIT7metadata:8 source: https://skillmd.com/skills/luokai0/oc-agent-browser-29 author: luokai010 version: "1.0"11 category: python-tools12---1314# Agent Browser1516You are an expert python engineer. Automates browser interactions for web testing, form.1718## Before Starting19201. **Goal** — what specific outcome do you need?212. **Environment** — versions, platform, existing setup?223. **Constraints** — performance, security, compatibility requirements?234. **Integration** — what systems does this connect to?245. **Output format** — code, config, script, or documentation?2526---2728## Core Expertise Areas2930- **Core implementation** — full working code for Agent Browser31- **Error handling** — robust error recovery and logging32- **Performance** — optimized patterns for production use33- **Testing** — unit and integration test strategies34- **Configuration** — environment-specific setup and tuning35- **Security** — secure coding patterns and best practices36- **Documentation** — clear API and usage documentation3738---3940## Key Patterns & Code4142### Core Implementation4344```python45import anthropic46import json4748client = anthropic.Anthropic()4950TOOLS = [51 {52 "name": "search",53 "description": "Search for information on a topic",54 "input_schema": {55 "type": "object",56 "properties": {57 "query": {"type": "string", "description": "Search query"}58 },59 "required": ["query"]60 }61 }62]6364def handle_tool(name: str, inputs: dict) -> str:65 if name == "search":66 return f"Search results for: {inputs['query']}"67 return f"Unknown tool: {name}"6869def run_agent_browser(task: str, max_turns: int = 10) -> str:70 messages = [{"role": "user", "content": task}]71 for _ in range(max_turns):72 response = client.messages.create(73 model="claude-sonnet-4-6",74 max_tokens=4096,75 tools=TOOLS,76 messages=messages77 )78 messages.append({"role": "assistant", "content": response.content})79 if response.stop_reason == "end_turn":80 return next((b.text for b in response.content if hasattr(b, 'text')), "")81 if response.stop_reason == "tool_use":82 results = [83 {"type": "tool_result", "tool_use_id": b.id,84 "content": handle_tool(b.name, b.input)}85 for b in response.content if b.type == "tool_use"86 ]87 messages.append({"role": "user", "content": results})88 return "Max turns reached"8990if __name__ == "__main__":91 result = run_agent_browser("AgentBrowser task: analyze and report")92 print(result)93```9495### Configuration & Setup96```python97# Agent Browser — Configuration98# Author: luo-kai (Lous Creations)99100config = {101 "name": "agent-browser",102 "version": "1.0.0",103 "author": "luo-kai",104 "enabled": True,105 "debug": False,106 "timeout_seconds": 30,107 "max_retries": 3,108}109```110111### Error Handling112```python113# Robust error handling pattern114import logging115logger = logging.getLogger("agent-browser")116117def safe_run(func, *args, **kwargs):118 try:119 return func(*args, **kwargs)120 except Exception as e:121 logger.error(f"agent-browser error: {e}", exc_info=True)122 raise123```124125---126127## Best Practices128129- **Fail fast with clear errors** — raise descriptive exceptions with context130- **Log at appropriate levels** — DEBUG for dev, INFO for ops, ERROR for problems131- **Validate inputs** — never trust external data without validation132- **Use type annotations** — improves IDE support and catches bugs early133- **Handle cleanup** — use context managers and `finally` blocks134- **Test edge cases** — empty inputs, nulls, max values, concurrent access135136---137138## Common Pitfalls139140| Pitfall | Problem | Fix |141|---------|---------|-----|142| No error handling | Silent failures in production | Wrap with try/except + logging |143| Hardcoded values | Not portable across environments | Use config/env vars |144| Missing timeouts | Hangs indefinitely | Always set timeout values |145| No retry logic | Single failure = broken workflow | Add exponential backoff |146| No cleanup on exit | Resource leaks | Use context managers |147148---149150## Related Skills151152- python-expert153- agent-browser-advanced154- performance-optimization155- error-handling156- testing-expert