Run deterministic multi-agent coding workflows with Bernstein
Orchestrate parallel CLI coding agents in isolated git worktrees with reproducible scheduling, gates, replay journals, and optional tamper-evident audit receipts.
npx skills add agentskillexchange/skills --skill run-deterministic-multi-agent-coding-workflows-with-bernstein
Use Bernstein when an engineering operator needs several CLI coding agents to work in parallel while preserving reviewability and reproducibility. The workflow decomposes a goal, dispatches tasks to supported agent CLIs in separate git worktrees, runs lint/type/test gates, records lineage and replay journals, and can emit HMAC-backed audit receipts for compliance-sensitive runs. This is not a generic agent framework: the boundary is deterministic CLI coding-agent orchestration, replay, lineage, and gated merge workflows inside repositories.
What this skill actually does
Inputs and prerequisites: Python 3.12+, pipx/pip/uv/Homebrew, git, and one or more supported CLI coding agents.
Setup notes: Install with pipx install bernstein, pip install bernstein, uv tool install bernstein, the published install script, or Homebrew, then run bernstein init followed by bernstein -g ‘ ‘ in a git repository.
Source and verification boundary: use https://bernstein.run as the canonical reference before running the workflow; keep commands, API calls, CLI usage, and generated outputs reviewable against that upstream source.
Framework fit: publish this as a Multi-Framework workflow only when the operator can invoke the documented toolchain directly, rather than treating the upstream project as a generic product listing.