claude-code-fusion

Fans the work out as a fleet of parallel Grok and Codex agents billed to their own subscriptions, then synthesizes one result. The peer engine equivalent of ultracode, adding intensity without spending Claude quota on the fleet. The default execution posture for any goal that decomposes into three or more independent work packages, not only an intensity booster for explicit asks.

Audits the orchestration setup without changes for model drift, stale pins, and environment overrides. Use when routing behaves oddly, after a CLI upgrade, or when model drift is suspected.

Aggregates delegation statistics and records semantic acceptance. Use after collecting and verifying a delegated result to record it as accepted or rejected, or when asked how the lanes are performing.

Runs a read only Codex review that challenges implementation and design decisions. Use when preparing to merge a substantial change or when a hostile second pass is wanted.

Cancels an active Codex job. Use when a running job must be stopped.

Lists canonical local Codex companion history with resumable thread IDs. Use when resuming or auditing past Codex work.

Forwards a request to the Codex rescue agent without exposing arguments to a shell. Use when Fusion routing or the user delegates work to Codex.

Shows or waits for a stored Codex job result. Use when collecting a background Codex job named by ID.

Runs a read only Codex review of local Git state. Use when the user wants Codex to review uncommitted or branch changes.

Checks whether the local Codex runtime is ready. Use when installing for the first time, after a CLI upgrade, or after a setup class failure.

Shows active and recent Codex jobs. Use when asked what Codex is doing or whether a job is still running.

Runs a Codex task through the companion runtime. Use when delegating a bounded implementation, investigation, or consultation to Codex.

Configures model defaults and capability scores across the Claude, Grok, and Codex orchestration surfaces. Use when changing model defaults, capability scores, or lane SKUs after the model lineup changes.

Convenes a blind Codex and Grok panel on one neutral brief and adjudicates the verdicts.

Installs or updates the orchestration rules in ~/.claude and checks the optional permission allow. Use on first install or after a plugin update changes the rules template.

Runs a Grok best-of-n implementation tournament and applies the winner. Use when several independent candidates and an automatic pick are wanted.

Cancel an active Grok job in this workspace. Use when a running job should stop.

Brief writing guidance for composing self contained Grok briefs for coding, review, diagnosis, and second opinion tasks

List canonical local Grok companion history and resumable sessions. Use when checking prior jobs or resumable sessions.

Show the stored result of a finished Grok job in this workspace. Use when retrieving a completed job's output.

Runs an adversarial Grok review of local git state. Use for a fast mechanical review sweep before merge, complementing deeper Codex review.

Check whether the local Grok CLI is ready for the companion runtime. Use when validating or changing the local companion setup.

Show aggregate stats for Grok jobs in this workspace. Use when checking Grok job activity over time.

Show active and recent Grok jobs for this workspace. Use when checking a job's progress or recent activity.

Delegates a coding or consultation task to the local Grok CLI. Use under a protected Grok role (burst, independence, live-web, large-context, best-of-n) or on explicit user request for Grok.

Internal contract for invoking the Codex companion runtime.

Internal helper contract for calling the grok-companion runtime from Claude Code

Internal guidance for presenting Grok companion output back to the user

Internal contract for returning Codex companion output without alteration.