nan023062/agent-team-cc

CBIM is a prompt/agent framework for Claude Code that enforces capability × business separation and adds a session-spanning memory pipeline. It structures knowledge as per-module .dna trees and ensures each task loads only the relevant agent “soul” × task subtree, reducing hallucination and keeping bounded context while preserving cross-session state. Who uses it: teams or projects that embed Claude Code into development workflows and need modular durable knowledge, reproducible agent behavior, and local tooling for memory/knowledge management. Key features (V1, available) - Kernel: prompts, agent definitions, Python hooks, managed venv, and a local MCP server. - Per-project install that drops a self-contained .cbim runtime (no global pip or CLI). - Core agents (architect, auditor, hr, programmer) and slash commands (/cbim_install, /cbim_help, /cbim_dashboard, /cbim_sched, ...). - MCP tools for memory and .dna inspection (memory_query, dna_list, agent_list, scheduler_status, etc.). - Embedded scheduler with persistent state and opt-in idle-aware tasks. - Project-local shims, isolated venv, .cbim/.venv, and gitignored runtime state.

AI-related keywords found in this repo's description, topics, or README summary — grouped by category. Each badge links to the corresponding ranking detail page.