菜单
Resolve NixOS issues using research and sequential thinking. Use when diagnosing or fixing problems on a NixOS host -- failed builds or switches, broken services or modules, configuration errors -- or whenever the user mentions NixOS, nixos-rebuild, or /etc/nixos.
默认使用会先检查来源的 Prompt；你也可以切换为直接命令，或下载本地副本。
决定是否安装前，请先阅读 SKILL.md，以及 SkillsMP 当前展示的配套文件。
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
直接命令不会经过审查 Prompt；运行前请先检查来源。
npx skills add https://github.com/jwiegley/nix-config --skill nixos命令会保持在同一行。复制前请横向滚动并检查完整内容。
想先保存到本地？可下载 SkillsMP 当前能够提供的文件。
正在显示 SKILL.md
| name | nixos |
| description | Resolve NixOS issues using research and sequential thinking. Use when diagnosing or fixing problems on a NixOS host -- failed builds or switches, broken services or modules, configuration errors -- or whenever the user mentions NixOS, nixos-rebuild, or /etc/nixos. |
Use nix-pro and the available NixOS tooling to resolve issues with the current NixOS installation.
Do not, under any circumstances, decrypt the SOPS secrets.yaml file. See the @CLAUDE.md file for extensive notes on this important security consideration.
Use available live web search as needed for research and discovering resources.
Use sequential-thinking when appropriate to break down tasks further.
On a managed NixOS host, run builds and switches through the host's build
driver from /etc/nixos, for example cd /etc/nixos && ./build switch.
For a custom command use ./build -- COMMAND .... The driver owns the
.nixos-build lock; never create, remove, or seize that path manually. If
the driver cannot acquire its lock, report its error and stop.
VPS is parked and must be selected explicitly for maintenance. On VPS, pass
--max-jobs 1 --cores 1 to every ./build build and ./build switch
invocation.
Methodology for the user-triggered /wiggum command (do not self-invoke). An autonomous-continuation loop for long-running work -- run, checkpoint, and verify until a defined Definition of Done holds or a stop-and-escalate condition fires. Covers durable handoff state, baseline re-verification after context compaction, per-commit self-audit, work-unit commit and restack cadence, subagent fan-out limits, and escalation.
Build, edit, and debug Node-RED flows on John's NixOS host (vulcan). Tuned to his actual plugin set, wiring conventions, naming style, and to the nodered_events PostgreSQL log + Grafana dashboard for chain tracing. Use whenever the user mentions Node-RED, flows.json, a flow tab name (Office, Schedule, Schedules, Pool Time, Away, Bedroom, TV Room, Institute Night, Debug), a Node-RED plugin or node type (chronos, api-call-service, api-current-state, server-state-changed, join-wait, actionflows, etc.), the Node-RED Events Grafana dashboard, or asks why a flow fired or didn't fire.
Offload safe, independent subtasks to concurrent subagents while continuing your own work as the coordinator, then integrate what they return. Use when mid-task and some work could run in parallel without conflicting -- research, generating a standalone new file, tests for a stable interface, docs, or isolated analysis -- to accelerate without races. The coordinator alone runs git and mutates shared state; subagents read freely but write only to isolated namespaces and hand back artifacts. Triggers include "parallelize this", "spin up subagents to help", "do these in parallel", "fan this out".
基于 SOC 职业分类
