Orchestrate a helioy-bus warroom: tmux agents doing parallel work under one orchestrator. Use for warroom, mixture of experts, MoE review, peer consensus, sign-off, brainstorm, spec-writing, scout, reuse audit, code-review, engineering, slice-build-loop, or any request that dispatches work to parallel agents.
Use for any helioy-bus mail operation: checking your inbox, sending messages to other agents, broadcasting to all agents, or responding to a "you have mail!" nudge. Use at the start of any task to check for directives from peer agents. Also use when the user says things like "reply to that agent", "tell nancy I'm done", "who's on the bus?", "who else is online?", "check for messages", "check for directives", or "send a message to X". Any inter-agent communication goes through this skill.
Improve codebase health through careful decomposition, consolidation, boundary repair, and developer experience cleanup. Use when the user asks for code hygiene, code-hygene, LOC reduction, refactoring, decomposing large files or functions, finding natural seams, reducing duplication, reorganizing modules, improving maintainability, or taking a craftsmanship pass across Rust, Python, TypeScript, JavaScript, Go, or any other language.
Generate accurate codebase onboarding maps and architecture diagrams. Use when the user asks for a codebase map, repo map, architecture map, onboarding map, dependency map, MAP.md, or a diagram that helps agents understand a codebase. For fmm indexed repos, use fmm MCP tools extensively and validate diagrams.
Claim a crates.io package name by checking availability and publishing a placeholder. Use when the user says "crate claim", "cargo claim", "reserve crate name", or wants to check if a crates.io name is available. Takes a crate name as an argument. Note: crates.io publishes are permanent (yank only, no unpublish).
Kubernetes fundamentals taught from first principles (kubernetes-the-hard-way backbone). Use BEFORE reasoning from scratch about any Kubernetes internals: cluster bootstrap, the control plane (apiserver / controller-manager / scheduler), etcd, the PKI/TLS cert mesh and CN/O-to-RBAC mapping, kubeconfigs, kubelet / containerd / CRI, pod networking and CNI, kube-proxy, encryption at rest, the Node Authorizer, or CRDs and the Helioy v2 K8s-shaped endgame. Routes you into the verified curriculum at ~/.mdx/knowledge/kubernetes/. Trigger when: explaining or debugging Kubernetes internals, designing K8s-shaped infrastructure, reviewing cluster config, or onboarding to how the core components fit together.
Use when planning, reviewing, or routing Linear work for Nancy or other autonomous agents. Covers issue capture, triage, planning gates, agent issue review, execution readiness, and Linear as the source of truth for autonomous work.
Check a package name across npm, PyPI, and crates.io, then scaffold a 0.0.1 placeholder for every available registry into one shared directory. Prints three copy-pasteable publish commands at the end so the user can run all three publishes themselves in their own terminal. Use when the user says "name claim", "claim X", "reserve name", or wants to lock a name across all three registries with a single command. Takes a name as an argument. Crates.io publishes are permanent (yank only, no unpublish).