---

name: meta-capabilities description: Advanced AI capabilities for paracle_meta including web search, code execution, MCP integration, and autonomous agent spawning. Use when you need enhanced AI functionality beyond basic generation. license: Apache-2.0 compatibility: Python 3.10+, requires configured capabilities metadata: author: paracle-team version: "1.0.0" category: integration level: expert display_name: "Meta Capabilities" tags: - capabilities - web - code-execution - mcp - spawning capabilities: - web-search - web-fetch - code-execution - mcp-tools - agent-spawning - filesystem - shell allowed-tools: Read Write Bash WebSearch WebFetch

Meta Capabilities Skill

Overview

This skill provides access to the advanced capability system of paracle_meta, enabling web research, code execution, MCP tool integration, and autonomous agent spawning.

Available Capabilities

Web Capabilities

Web Search

Copyfrom paracle_meta import MetaAgent

async with MetaAgent() as meta:
    # Search the web
    results = await meta.web_search(
        "Python async best practices 2025",
        max_results=10
    )

    for result in results:
        print(f"{result.title}: {result.url}")

Fetch Web Content

Copy# Fetch and parse a page
content = await meta.web_fetch(
    "https://docs.python.org/3/library/asyncio.html",
    extract_text=True
)

# Get specific content
code_examples = await meta.web_fetch(
    url,
    selector="pre.python"  # CSS selector
)

Code Execution

Safe Execution

Copy# Execute Python code safely
result = await meta.run_code(
    code="""
import math
result = math.factorial(10)
print(f"10! = {result}")
""",
    language="python",
    timeout=30  # seconds
)

print(result.stdout)  # "10! = 3628800"
print(result.exit_code)  # 0

Sandboxed Execution

Copyfrom paracle_meta import CodeExecutionCapability, CodeExecutionConfig

config = CodeExecutionConfig(
    sandbox=True,
    max_memory_mb=512,
    max_cpu_seconds=30,
    allowed_imports=["math", "json", "datetime"]
)

capability = CodeExecutionCapability(config)
result = await capability.execute(code, "python")

MCP Integration

Using MCP Tools

Copyfrom paracle_meta import MCPCapability, MCPConfig

# Connect to MCP server
config = MCPConfig(
    server_url="http://localhost:3000",
    tools=["read_file", "write_file", "search"]
)

mcp = MCPCapability(config)
await mcp.initialize()

# Call a tool
result = await mcp.call_tool(
    "read_file",
    {"path": "/path/to/file.py"}
)

Discovering Tools

Copy# List available tools
tools = await mcp.list_tools()
for tool in tools:
    print(f"{tool.name}: {tool.description}")

Agent Spawning

Spawn Specialized Agents

Copyfrom paracle_meta import AgentSpawner, SpawnConfig

spawner = AgentSpawner()

# Spawn a researcher agent
researcher = await spawner.spawn(
    SpawnConfig(
        name="Researcher",
        agent_type="researcher",
        task="Research Python testing frameworks",
        capabilities=["web_search", "web_fetch"]
    )
)

# Wait for results
result = await researcher.wait()
print(result.findings)

Parallel Agents

Copy# Spawn multiple agents for parallel work
agents = await spawner.spawn_many([
    SpawnConfig(name="SecurityReviewer", task="Review for security issues"),
    SpawnConfig(name="PerformanceAnalyzer", task="Analyze performance"),
    SpawnConfig(name="DocumentationChecker", task="Check documentation"),
])

# Gather results
results = await asyncio.gather(*[a.wait() for a in agents])

File System Operations

Read/Write Files

Copyfrom paracle_meta import FileSystemCapability

fs = FileSystemCapability()

# Read a file
content = await fs.read("src/main.py")

# Write a file
await fs.write("output/report.md", report_content)

# Search files
matches = await fs.search(
    pattern="def test_",
    path="tests/",
    file_types=["*.py"]
)

Git Operations

Copy# Git status
status = await fs.git_status()

# Git diff
diff = await fs.git_diff("HEAD~1")

Shell Capability

Execute Commands

Copyfrom paracle_meta import ShellCapability, ShellConfig

shell = ShellCapability(ShellConfig(
    allowed_commands=["ls", "cat", "grep", "find"],
    blocked_patterns=["rm -rf", "sudo"]
))

result = await shell.execute("ls -la src/")
print(result.stdout)

Memory Capability

Persistent Context

Copyfrom paracle_meta import MemoryCapability

memory = MemoryCapability()

# Store information
await memory.remember("api_key", "sk-...", ttl=3600)
await memory.remember("user_preferences", {"theme": "dark"})

# Recall information
api_key = await memory.recall("api_key")
prefs = await memory.recall("user_preferences")

# Search memory
results = await memory.search("api")

Provider Chain

Fallback Strategy

Copyfrom paracle_meta import ProviderChain, FallbackStrategy

chain = ProviderChain(
    providers=["anthropic", "openai", "ollama"],
    strategy=FallbackStrategy.SEQUENTIAL
)

# Automatically falls back on failure
result = await chain.complete(messages)

Circuit Breaker

Copyfrom paracle_meta import CircuitBreaker

breaker = CircuitBreaker(
    failure_threshold=3,
    recovery_timeout=60
)

# Prevents cascading failures
result = await breaker.call(provider.complete, messages)

Configuration

Environment Variables

Copy# Web capabilities
PARACLE_WEB_TIMEOUT=30
PARACLE_WEB_MAX_RESULTS=10

# Code execution
PARACLE_CODE_SANDBOX=true
PARACLE_CODE_MAX_MEMORY=512

# MCP
PARACLE_MCP_SERVER_URL=http://localhost:3000

# Spawning
PARACLE_MAX_SPAWNED_AGENTS=10

Security Considerations

1. Code Execution: Always use sandbox mode in production
2. Shell Commands: Whitelist allowed commands
3. Web Access: Respect rate limits and robots.txt
4. Agent Spawning: Limit concurrent agents
5. Memory: Don't store secrets in plain text

Related Skills

• meta-generation: Generate artifacts
• meta-learning: Improve over time
• tool-integration: Create custom tools