novotnyllc/railyard

cleanup-codex

Clean completed-session process residue, inspect macOS Codex app-server resources, reap exact snapshot-bound residue, or explicitly recycle one fully attested detached server.

deliver

Route one host-local software change or pull-request task through Railyard. Implementation/ship requests MUST dispatch an LFG delivery carrier before implementation or remote shipping; this coordinator never substitutes direct implementation for the LFG…

oracle

Oracle second-model review: bundle prompts/files, debug, refactor, design-check.

thermo-nuclear-review

Comprehensive security and correctness audit of a branch's changes. Use for thermo nuclear, thermonuclear, or deep review requests, or branch/PR diff audits focused on bugs, breaking changes, security issues, devex regressions, and feature-gate leaks.

orchestrate

Orchestrate configured fleet/account delivery or complex objectives across independently resumable tasks, projects, hosts, pull requests, and dependencies while remaining available as the control task. Use when configured routing owns allocation, when the…

model-routing

Resolve one bounded model, effort, transport, and budget decision through railyard/model-routing/v1. Use before model-specific task, subagent, provider-review, or steering actions.

setup

Set up, extend, or diagnose the railyard delivery system: inventories what is installed, installs prerequisites (Compound Engineering, ponytail, roundhouse, agent-utilities, gh-stack, tmux/jq) with one consent, asks what the fleet hosts are and enrolls each…

audit

Reconstruct how a chunk of agent work actually ran — which skills routed, what decided what, how many subagents fanned out on which models and why, retries, review rounds, and whether it matched the planned shape. Use when the user asks how a run worked, to…

doctor

Diagnose and fix the delivery system: plugin and skill version sync across harnesses and fleet hosts, marketplace freshness, fleet config validity, host reachability and enrollment state, executor integrity, and Compound Engineering compatibility. Use when…

thermos

Launch both thermo-nuclear review subagents in parallel, then synthesize their findings. Use for thermos, double thermo review, or combined bug/security and code-quality branch audits.