By mike-bronner
Calvinball-driven development pipeline: Miss Wormwood (triage), Moe (dev work), and Tracer Bullet (code review). A local scheduled Haiku task called Dispatch polls Calvinball every 20–30 minutes via MCP and fires the right agent for each pending item.
npx claudepluginhub mike-bronner/claude-workbench --plugin workbench-dev-teamTriage agent. Dispatched by Dispatch (the orchestrator) on unrefined GitHub project items. Inspects the issue + repo, generates acceptance criteria, scores WSJF fields, and moves the item to Backlog.
Development agent. Two operating modes detected from input shape — Calvinball mode (when invoked with an item ID, runs the full pipeline orchestration: lock, fetch state, branch, draft PR, status transitions, cleanup) and Direct mode (when invoked with prose, runs the universal dev workflow with no Calvinball calls — intended for ad-hoc dev work delegated from Claude Code or Cowork). In both modes, the actual coding follows the /workbench-dev-team:develop skill — that skill is the canonical source of truth for development standards.
Code review agent. Dispatched by Dispatch (the orchestrator) on items in "In Review" status. Finds the associated PR, checks against acceptance criteria, approves or requests changes, and escalates to Mike after 3 rounds.
Apply universal development standards when implementing code changes, fixing bugs, refactoring, writing tests, or any task that writes or modifies source code. Use this skill BEFORE writing or changing code — every time, manual or agent-driven — to ensure consistent conventions, testing, commit hygiene, and a human-in-the-loop decision protocol across all development work. Triggers on requests to "implement", "build", "fix", "add", "refactor", "write a test", "code up", or any prompt that produces code changes.
Generate commit messages using Conventional Commits + Gitmoji format. Use this skill whenever creating, drafting, or suggesting git commit messages — including /commit commands, pre-commit hooks, bulk commits, and any context where a commit message is being composed. Always invoke this skill before writing a commit message.
Autonomous multi-agent development pipeline — from ticket to PR in one command
Uses power tools
Uses Bash, Write, or Edit tools
Share bugs, ideas, or general feedback.
24 commands, 8 agents. Loops, swarms, and teams powered by Claude Code's built-in Task System. Native task dependencies, ctrl+t progress, automatic persistence.
Long-running agent harness with 5-layer memory architecture, GitHub integration, autonomous batch processing, Agent Teams with ATDD, 9 hooks (safety, quality gates, team coordination), and 6 Agent Skills
Providers (GitHub/GitLab/Jira/ClickUp/Slack/Fireflies/Confluence), MCP server, and 24 skills with token-aware output compression.
Multi-agent orchestration framework for Claude Code, Gemini CLI, and Codex CLI — 19 agents, 13 skills, 15 commands, quality gates, TDD enforcement
127-agent automated development system with Agent Teams, quality gates, Bug Council diagnostics, and autonomous execution