A Team — A Complete Engineering Team in One Folder v1.4.1

Not a marketplace of agents you configure — a pre-configured, pre-enforced team of 26 specialists with a lead orchestrator, hard quality gates, and a Pipeline Auditor that verifies work was done, not just reported. Drop this folder into any project and it's operational from the first keystroke. Works on Claude Code, Codex CLI, Cursor, OpenCode, and GitHub Copilot CLI.

Quick Start

1. Install the shared directories, scripts and platform integration files using install.sh / install.ps1, or follow the complete manual options in README.md. Do not overwrite an existing AGENTS.md, CLAUDE.md or .claude/settings.json.
2. Write a project INIT.md (use INIT_TEMPLATE.md as a guide).
3. Run /orchestrate init — the Lead Orchestrator reads INIT.md, evaluates which agents and skills are irrelevant to the scope, and prunes them automatically.
4. Every session starts with the using-a-team meta-skill injected via the SessionStart hook.

Folder Layout

A Team/
├── CLAUDE.md                  ← This file
├── AGENTS.md                  ← Agent roster & quick reference
├── INIT_TEMPLATE.md           ← Template for project-specific init doc
├── .claude/
│   ├── settings.json          ← Universal permissions, hooks, MCP stubs
│   ├── agents/                ← 26 agent profiles
│   ├── commands/              ← Slash commands
│   └── rules/                 ← Enforced coding & workflow standards
├── skills/                    ← 20 workflow skill modules
├── hooks/
│   └── session-start.md       ← Injected at every session start
├── .cursor-plugin/            ← Cursor IDE integration
├── .codex-plugin/             ← Codex CLI integration
├── .copilot-plugin/           ← GitHub Copilot CLI integration
└── .opencode/                 ← OpenCode integration

Agent Roster (26)

Skill Library (20)

Orchestrator Init & Pruning Protocol

When you run /orchestrate init:

1. The Lead Orchestrator reads INIT.md (your project scope document).
2. It evaluates each agent and skill against the declared scope.
3. It deletes files that are irrelevant (e.g., removes go-reviewer.md from a Python-only project).
4. It generates .agent-sync/TEAM.md — permanent record of active agents.
5. It generates .agent-sync/ROUTING.md — task routing table for this project.

The pruned workspace is leaner and agents no longer have access to context that doesn't apply.

Session Enforcement

Every session start triggers the hooks/session-start.md content via the SessionStart hook. This injects the using-a-team meta-skill into session context, which enforces:

• Skills must be consulted before action
• verification-before-completion before any "done" claim
• code-reviewer after every code change
• Language-specific reviewers for Go/Python/Rust changes

Standards

All agents enforce the rules in .claude/rules/:

• orchestration.md — Orchestrator state machine, init/prune, veto buffer protocol
• coding-style.md — Immutability, KISS, DRY, YAGNI, naming conventions
• security.md — Secrets, injection, auth, OWASP mandatory checks
• testing.md — TDD, 80% coverage, unit + integration + E2E
• git-workflow.md — Conventional commits, PR process
• performance.md — Model selection, context window management
• agents.md — Parallel execution, agent selection rules
• patterns.md — Repository pattern, API envelopes, error handling

Cross-Platform Support