This skill should be used whenever the user asks to "make a wiring diagram", "update the wiring diagram", "fix the diagram", "show me how to wire this", "draw a hookup guide", or otherwise wants a picture of how components connect for an ESP32/Pi/Arduino/discrete-component build. It generates a PICTORIAL breadboard-style diagram — real Fritzing part photos + clean colored point-to-point wires, rendered offline and headless from a small JSON spec — NOT a schematic, because the reader cannot read schematics. It must not be used for PCB layout or schematic capture (that is Eagle/KiCad).
Operate tmux as an agentic multiplexer on a live, shared machine — run and coordinate parallel background work without disrupting the user's or other agents' sessions. Use when driving tmux non-interactively (new-session -d, send-keys, kill-session) to run long-running processes, orchestrate parallel jobs, or coordinate between agents, ESPECIALLY where the user or other agents already have tmux sessions. NOT for configuring tmux (.tmux.conf / keybindings). Triggers on tmux new-session, send-keys, kill-session, "run it in tmux", "background job in tmux", parallel jobs, driving tmux on a shared/remote box.
Set up and use the OpenCode "zen" model harness in any project — route agent work to cheaper models by capability (fast / reasoning-heavy / code-generation) to save Claude's token budget, with Claude orchestrating and validating. Use when integrating OpenCode, wiring kimi-k2.6 / glm-5.2, creating ~/.config/vinay-agent/config.toml, offloading reasoning or code generation to a cheaper endpoint, or deciding which model a task should go to. Triggers on "opencode", "opencode harness", "vinay agent", "route to a cheap model", "save tokens", "kimi", "glm", "/opencode-harness".
Architecture guidance for building and maintaining cross-platform Expo apps (iOS, Android, web) from one codebase — the taste that keeps a growing app factored without over-building it. Use when building or extending an Expo / React Native app, setting up an Expo + Next.js monorepo, writing expo-router navigation, sharing code across native and web, styling with NativeWind, wrapping native SDKs, or configuring EAS Build/Update. Triggers on "expo app", "react native app", "build a mobile app", "ios android web", "expo router", "expo monorepo", "eas build", "nativewind".
Build, author, or improve a Claude skill that transfers real domain expertise. Use when creating a skill from authoritative docs, distilling documentation into LLM-digestible guidance, deciding what belongs in a skill, or proving a skill actually improves output. Covers the gather-distill-gap-detect-hill-climb-validate loop and the 4-cell benchmark that proves a skill transfers expert taste to a non-expert. Triggers on "build a skill", "make a skill", "improve my skill", "is this skill any good", "turn these docs into a skill".
Embedded systems design for ESP32/Arduino. Use when writing C/C++ for ESP32, designing firmware architecture, managing peripheral interactions, or optimizing for resource constraints. Focuses on the two things Claude reliably slips on -- ISR primitives and proving the design with measurement -- not on re-teaching architecture it already does well.
End-to-end setup guide for claude-peers -- broker vs client decision, UCAN token exchange, per-project agent identity files, push vs polling delivery.
Design thinking for building systems with Claude Code. Use when building hooks, daemons, agent pipelines, skills, multi-agent systems, background automations, or any system where Claude Code components connect. Also use when the task involves autonomous agents, scheduled tasks, or any pipeline where LLM output feeds into another system.