Orchestration

Frame a decision as a multi-model debate and launch it with a clear motion

Scope requirements by consulting several AI models in a define phase

Validate, score and review work with several AI providers in one pass

Implement changes with several external AI providers in a develop phase

Research a topic across several external AI providers in a discover phase

Split large migrations and multi-issue fixes into parallel tasks with quality gates

Handle small ad-hoc tasks quickly without the overhead of a full workflow

Check whether a multi-agent setup earns its coordination cost before adding agents

Detect whether a task is coding or knowledge work to pick the right workflow

Add GitHub Copilot CLI as a zero-cost model provider in multi-model workflows

Project remaining workflow cost from phase averages and warn before budget overruns

Run a multi-LLM council with personas, budget caps and a synthesized verdict

Run structured debates across several AI providers for high-stakes decisions

Run a hands-off build-and-ship pipeline that turns a spec into a project

Repeat a task in a loop, polling or refining until the goal converges

Decide when init, review and security requests need multi-model routing

Split migrations and multi-file refactors across parallel agents

Show where you are in the workflow and what to do next

Delegate independent analysis work to worker agents with self-contained briefs

Set an explicit objective with success criteria and stopping conditions

Run planning, review, compare and export workflows as state machines over sessions

Stop active turns, compute jobs and kernels in the current session

Inspect and manage bb threads, projects, machines, providers and plugins

Delegate a self-contained coding task to Codex CLI as a subagent

Draft goal prompts and set up persistent autonomous agent loops

Inspect, message and coordinate agents running in Herdr workspaces

Launch a new bb worker thread with its project, model, worktree and brief

Archive the current Cloudroom thread and release its runtime

Split long autonomous tasks into gated steps and verify evidence before reporting

Get a second opinion on changes from Google Gemini in review or challenge mode

Get a second-model review by bundling prompts and files for debugging or design

Request a second-model consultation over API, browser or render with run provenance

Run a named team of cooperating Codex workers with durable, saved state

Apply the ultrawork directive for full task execution in omo on Codex

Execute a written work plan with checkpoints, evidence logs and parallel subagents

Break goals into evidence-bound steps in a durable, checkpointed loop

Apply the seek-truth-from-facts principle and route tasks to the right method skill

Pick one main target when many tasks compete, finish it and verify before switching

Test a hypothesis in practice and iterate in a loop with a stop condition

Chain several thinking methods into standard patterns with handoffs and stop rules

Coordinate research, strategy and finance agents into a GO or NO-GO briefing

Handle a change request from another session, confirming scope before editing

Run long or flaky tasks as bounded, resumable loops with saved checkpoints

Plan, delegate and review coding agent work with scope and acceptance checks

Ask a session in another repo for a change, with enough detail to need no follow-ups

Hand tasks to Codex or Gemini CLI for a second opinion and critique the result

Design multi-agent team definitions in the open .envelope.json format

Schedule a future resume of GSD work at a set time or interval

Turn a vague goal into a verifiable completion contract, then run it with a cap