Summary

This repository provides the Flowbaby Agent Team, a multi-agent workflow system for VS Code designed to bring structure, quality gates, and persistent long-term memory to AI-assisted development. It addresses the chaotic nature of current AI coding assistants by offering specialized agents that manage distinct workflow phases, ensuring context retention and auditable processes. The system benefits developers seeking a more organized, reliable, and quality-assured AI development experience.

How It Works

The core approach employs specialized AI agents, each responsible for a specific development task (e.g., planning, coding, security review), enforcing strict separation of concerns. Agents produce structured Markdown documents, creating an auditable trail of decisions and actions. Crucially, the system integrates with the Flowbaby VS Code extension, providing a persistent memory layer that allows agents to retain context and decisions across sessions, treating memory as core infrastructure. This design overcomes limitations like context loss and lack of process rigor inherent in simpler AI assistants.

Quick Start & Requirements

• Install: Clone the repository (git clone https://github.com/groupzer0/agents.git) and copy desired agent .agent.md files into your project's .github/agents/ directory. Alternatively, install agents at the user level via the VS Code Command Palette (Chat: New Custom Agent → select User profile).
• Run: Select the agent from the VS Code Copilot Chat dropdown and provide your prompt.
• Prerequisites: VS Code with GitHub Copilot, Flowbaby VS Code extension (for memory), Python 3.10+.
• Links: Flowbaby VS Code Marketplace: https://marketplace.visualstudio.com/items?itemName=flowbaby.flowbaby. Comprehensive documentation is available within the repository (USING-AGENTS.md, AGENTS-DEEP-DIVE.md).

Highlighted Details

• Separation of Concerns: Dedicated agents for distinct workflow phases (Roadmap, Planner, Implementer, Security, etc.) prevent scope creep.
• Document-Driven Workflow: All agent interactions and decisions are recorded in structured Markdown documents, forming an audit trail.
• Integrated Quality Gates: Agents like Critic, Security, and Code Reviewer enforce quality checks at various stages.
• Persistent Memory: Flowbaby integration ensures agents retain context and decisions across development sessions.
• Comprehensive Security Agent: Utilizes a Five-Phase Framework (Architectural, Code, Dependency, Infrastructure, Compliance) for deep security reviews, covering OWASP Top 10, CVEs, and more.
• Skills System: Agents leverage modular, reusable instruction sets (Skills) for on-demand capabilities like memory management, security patterns, and engineering standards.

Maintenance & Community

Contributions are welcomed, with specific areas of interest noted. The repository includes automated Markdown linting via GitHub Actions. No direct links to community channels (Discord/Slack), roadmaps, or specific contributor/sponsorship information were found in the README.

Licensing & Compatibility

• License: MIT License.
• Compatibility: The MIT license permits commercial use and integration with closed-source projects without significant restrictions.

Limitations & Caveats

The full value of the agents is realized only when using the Flowbaby extension for persistent memory; without it, agents operate stateless. A known upstream bug in the GitHub Copilot CLI prevents user-level agents from loading, necessitating the recommended per-repository agent setup (.github/agents/) until the issue is resolved. The project appears under active development, with recent updates detailing significant changes to its skill system and agent workflows.