A lightweight VS Code agent plugin for role-separated AI development. It provides a Producer, a stack-adaptive Dev Team, optional QA, and concise templates for planning, brainstorming, and context recovery.

Nova, Sage, and Milo are adaptive perspectives inside the Dev agent, not mandatory project layers or separate sessions.

The ai-team skill includes:

• a concise project brief template;
• a proportional work-plan template;
• brainstorm guidance with distinct perspectives;
• practical anti-patterns learned from multi-agent projects.

Plan -> Implement -> Test -> optional review or QA -> Merge -> update project state

Use the lightest process that fits the work:

• proceed directly for small, clear changes;
• create a short plan for substantial or cross-cutting work;
• add independent review or QA when risk, uncertainty, or repository policy warrants it;
• rely on repository branch protection, required checks, permissions, and merge policy.

1. Run Chat: Install Plugin From Source in VS Code.
2. Enter https://github.com/denis-a-evdokimov/ai-team-orchestration.git.
3. Select one of the bundled agents in Chat.

"chat.pluginLocations": {
  "/path/to/ai-team-orchestration": true
}

@ai-team-producer Help me define the next useful outcome for this project.
Use the ai-team skill. Keep planning proportional to the work.

@ai-team-dev Read the repository instructions and active plan or issue.
Implement the requested outcome, run relevant checks, self-review the diff,
and prepare the pull request. Do not merge.

@ai-team-qa Test the pull request against its acceptance criteria.
Focus on important behavior and regressions. Report Ready, Ready with minor
follow-ups, or Blocked with reproducible findings.

The bundled agents intentionally omit tools and model frontmatter.

• Built-in, MCP, and extension-provided tools enabled by the developer remain available.
• The developer keeps control of model selection.
• Agent roles are enforced by instructions plus normal trust, authentication, permission, and approval controls.

If too many tools are enabled, deselect irrelevant tools or MCP servers, or configure github.copilot.chat.virtualTools.threshold. Do not copy a machine-specific tool list into the plugin.

Keep the stable agent IDs:

• ai-team-producer
• ai-team-dev
• ai-team-qa

Adapt role personalities and project perspectives as needed. QA and independent review are optional unless repository policy or risk makes them necessary.

Version 2:

• keeps the stable ai-team-* agent IDs;
• inherits the user's enabled tools and selected model;
• makes roles stack-adaptive;
• replaces the formal delivery protocol with a lightweight, risk-proportionate default workflow.

The standalone skill remains ai-team. The Awesome Copilot export adapts only its frontmatter name to ai-team-orchestration.

This repository is canonical. See AGENTS.md for ownership rules and CONTRIBUTING.md for validation and Awesome Copilot export instructions.

npm run validate
npm test
npm run awesome:check -- --target <awesome-copilot-checkout>
npm run awesome:prepare -- --target <awesome-copilot-checkout> --output <patch-file>

The original workflow grew from Arcade After Dark, a multi-agent project built with distinct planning, implementation, design, and QA perspectives.

MIT