ai-provenance/devmemory-coordination

Universal Agent Coordination System

This skill works with ANY AI coding agent (Cursor, Claude, Copilot, Mistral, Antigravity, etc.)

All agents share the same memory system. What one agent learns, all agents can use.

1. Memory Access Pattern (Universal)

# Get complete context for your task
from devmemory.agent_tools import get_universal_agent_tools
memory = get_universal_agent_tools()
context = memory.get_hierarchical_context("implement authentication system")

2. Cross-Agent Learning

# Store what you've learned (universal format)
memory.store_agent_learning(
    learning="Use OAuth2 with PKCE for mobile auth to prevent token theft",
    learning_type="semantic",
    topics=["security", "authentication", "mobile"],
    entities=["OAuth2", "PKCE"]
)

3. Reusable Skills System

# Check if skill exists
skill = memory.get_agent_skill("authentication_pattern")

if not skill:
    # Store new skill for all agents to use
    memory.store_agent_skill(
        skill_name="authentication_pattern",
        skill_description="Secure authentication pattern using JWT with refresh tokens",
        implementation="1. Use short-lived access tokens...",
        use_cases=["Web applications", "Mobile apps"]
    )

4. Coordination Protocol

# Check who's working on what
coordination = memory.get_hierarchical_context("current work")
active_sessions = coordination["coordination"]["active_sessions"]

# Announce your work
memory.store_agent_learning(
    learning="Currently implementing OAuth2 provider integration",
    topics=["active-work", "coordination"],
    entities=["authentication"]
)

Keep memories relevant

Store:

• Architecture decisions with rationale
• Patterns and conventions
• Gotchas and workarounds
• API quirks and limitations
• Performance optimizations

Avoid:

• Implementation details obvious from code
• Temporary debugging notes
• Personal preferences
• Redundant copies of commit messages