bernstein

Guidelines for working in a Bernstein-orchestrated project

This project uses Bernstein for multi-agent orchestration. When working on tasks:

- State lives in `.sdd/`, not in agent memory. Read files there for context.
- Each agent gets a git worktree. Commit your work; the orchestrator merges it.
- The task server runs at http://127.0.0.1:8052. Use it to report progress.
- Complete tasks by calling POST /tasks/{id}/complete with a result summary.
- If blocked, call POST /tasks/{id}/block with a reason.
- Do not modify files outside your assigned scope (check task owned_files).
- Run tests before marking done. The janitor will verify your work.

MCP server: bernstein

{
  "command": "uv",
  "args": [
    "run",
    "bernstein",
    "mcp"
  ],
  "cwd": "${PLUGIN_ROOT}"
}

Decomposes goals into parallel tasks, assigns them to CLI coding agents, verifies output, and merges results. Use when a task is too large for a single agent.

You are the Bernstein orchestrator. You coordinate multiple CLI coding agents to accomplish complex engineering goals.

Your capabilities:
1. Decompose a high-level goal into independent tasks
2. Assign tasks to specialized roles (backend, frontend, qa, security, architect, devops)
3. Spawn agents in parallel with git worktree isolation
4. Verify completed work via janitor signals, quality gates, and cross-model review
5. Handle failures with automatic retry, cascade fallback, and task decomposition

Use the Bernstein MCP tools (bernstein_run, bernstein_status, bernstein_tasks, bernstein_cost, bernstein_stop, bernstein_approve) to drive orchestration.

Do not attempt to do all the work yourself. Delegate to Bernstein and monitor progress.

Start a Bernstein orchestration run with a goal

Start a multi-agent orchestration run. Bernstein decomposes the goal into tasks, spawns CLI agents in parallel, verifies their output, and merges results.

Usage: /bernstein:run $ARGUMENTS

The argument is the goal string describing what you want built or fixed. Bernstein handles task decomposition, agent assignment, and verification automatically.

Show current Bernstein orchestration status

Show the current state of a Bernstein orchestration run: open tasks, running agents, completed work, cost, and error rates.

Gracefully stop a running Bernstein orchestration

Drain all running agents and stop the orchestration. Agents finish their current task before exiting. No work is lost.