$npx -y skills add dsifry/metaswarm --skill startUse when starting work on any task, when the user mentions metaswarm, or when the user wants to begin tracked development work
| 1 | # BEADS Multi-Agent Orchestration Skill |
| 2 | |
| 3 | This skill coordinates a swarm of specialized AI agents to autonomously handle GitHub Issues from creation to merged PR. |
| 4 | |
| 5 | --- |
| 6 | |
| 7 | ## Quick Start |
| 8 | |
| 9 | ### Start Work on a GitHub Issue |
| 10 | |
| 11 | ```bash |
| 12 | # User triggers via any of: |
| 13 | @beads start #123 |
| 14 | bd start 123 |
| 15 | /beads-start 123 |
| 16 | ``` |
| 17 | |
| 18 | ### Check BEADS Status |
| 19 | |
| 20 | ```bash |
| 21 | bd ready # Show tasks ready to work |
| 22 | bd list # Show all tasks |
| 23 | bd stats # Show project statistics |
| 24 | bd doctor # Check system health |
| 25 | ``` |
| 26 | |
| 27 | --- |
| 28 | |
| 29 | ## Agent Roster |
| 30 | |
| 31 | | Agent | Role | Spawned When | |
| 32 | | ------------------------- | ------------------------------ | ---------------------------------- | |
| 33 | | **Issue Orchestrator** | Main coordinator per Issue | Issue receives `agent-ready` label | |
| 34 | | **Researcher Agent** | Codebase exploration | Orchestrator creates research task | |
| 35 | | **Architect Agent** | Implementation planning | Research complete | |
| 36 | | **Product Manager Agent** | Use case & user benefit review | Design review gate (parallel) | |
| 37 | | **Designer Agent** | UX/API design review | Design review gate (parallel) | |
| 38 | | **Security Design Agent** | Security threat modeling | Design review gate (parallel) | |
| 39 | | **CTO Agent** | TDD readiness & plan review | Design review gate (parallel) | |
| 40 | | **Coder Agent** | TDD implementation | Design review gate approved | |
| 41 | | **Code Review Agent** | Internal code review | Implementation complete | |
| 42 | | **Security Auditor** | Security review (code) | Implementation complete | |
| 43 | | **Release Engineer Agent** | Safe delivery from merge through production | QA approves PR, PR reaches merge readiness | |
| 44 | | **PR Shepherd** | PR lifecycle management | PR created | |
| 45 | |
| 46 | See `./agents/` directory for detailed agent definitions. |
| 47 | |
| 48 | --- |
| 49 | |
| 50 | ## Design Review Gate (NEW) |
| 51 | |
| 52 | For complex features created via brainstorming, an automatic **Design Review Gate** ensures quality before implementation: |
| 53 | |
| 54 | ``` |
| 55 | Design Document Created |
| 56 | │ |
| 57 | ▼ |
| 58 | ┌─────────────────────────────────────────────────┐ |
| 59 | │ DESIGN REVIEW GATE │ |
| 60 | │ │ |
| 61 | │ Spawns in PARALLEL: │ |
| 62 | │ • Architect Agent (technical architecture) │ |
| 63 | │ • Designer Agent (UX/API design) │ |
| 64 | │ • UX Reviewer (user flows, integration WUs) │ |
| 65 | │ • CTO Agent (TDD readiness) │ |
| 66 | │ │ |
| 67 | │ ALL must approve to proceed │ |
| 68 | └─────────────────────────────────────────────────┘ |
| 69 | │ |
| 70 | ├── Any NEEDS_REVISION? → Iterate on design (max 3x) |
| 71 | │ |
| 72 | ALL APPROVED |
| 73 | │ |
| 74 | ▼ |
| 75 | Create BEADS Epic → Begin Implementation |
| 76 | ``` |
| 77 | |
| 78 | ### Triggering the Design Review Gate |
| 79 | |
| 80 | The gate is automatically triggered when: |
| 81 | |
| 82 | - `superpowers:brainstorming` completes and commits a design doc |
| 83 | - User runs `/review-design <path-to-design.md>` |
| 84 | |
| 85 | ### Review Criteria by Agent |
| 86 | |
| 87 | | Agent | Focus Areas | |
| 88 | | --------------- | ---------------------------------------------------------- | |
| 89 | | Product Manager | Use case clarity, user benefits, scope, success metrics | |
| 90 | | Architect | Service architecture, dependencies, patterns, integration | |
| 91 | | Designer | API design, UX flows, developer experience, consistency | |
| 92 | | Security Design | Threat modeling, auth/authz, data protection, OWASP Top 10 | |
| 93 | | UX Reviewer | User flows, text wireframes, integration WUs, empty/error states | |
| 94 | | CTO | TDD readiness, codebase alignment, completeness, risks | |
| 95 | |
| 96 | ### Iteration Protocol |
| 97 | |
| 98 | - **Max 3 iterations** before human escalation |
| 99 | - Each iteration: revise design → re-run all reviewers |
| 100 | - Escalation options: Override / Defer / Cancel |
| 101 | |
| 102 | See the `design-review-gate` skill for full details. |
| 103 | |
| 104 | --- |
| 105 | |
| 106 | ## Team Mode Coordination |
| 107 | |
| 108 | When multiple Claude Code sessions are active on the same repository (e.g., parallel worktrees), metaswarm automatically enters **Team Mode**. In Team Mode, agents behave as persistent teammates with context retention across sessions and direct inter-agent messaging for coordination. Mode detection is automatic based on the presence of concurrent sessions. |
| 109 | |
| 110 | For the full Team Mode protocol — including message routing, context sharing, and conflict resolution — see `./guides/agent-coordination.md`. |
| 111 | |
| 112 | --- |
| 113 | |
| 114 | ## Plan Review Gate |
| 115 | |
| 116 | After the Architect creates an implementation plan and bef |