Claude Autoshutdown — Development skill for Claude Code
Shut down Windows only after every Claude Code session and its subagents have really finished.
How to install Claude Autoshutdown
This entry records only its repository, not the path inside it, so there is no
exact command to give. Open kamiljan11/claude-autoshutdown and copy the folder into
~/.claude/skills/, or the file into ~/.claude/agents/.
What Claude Autoshutdown does
Shut down Windows only after every Claude Code session and its subagents have really finished. Reads agent turn state instead of guessing from CPU, 14 safety gates, no UAC prompt. Python + Tkinter, zero runtime dependencies.
Alternatives in Development
- Investigate Performance — Investigate a Nimbalyst performance problem (freeze, lag, idle CPU, slow op, memory growth) with measured evid 1.6k ★
- LiquidGlassReference — iOS 26 Liquid Glass: Ultimate Swift/SwiftUI Reference 266 ★
- Auto Continue — You are the auto-continue gate for Omux's Main Agent 94 ★
README
Claude AutoShutdown — shut down Windows only after every Claude Code session has really finished
Leave your AI agents running overnight and let the machine turn itself off — but only once **every** Claude Code / Cowork session, including its subagents, has genuinely stopped working.
A small Windows desktop app (Python + Tkinter, **no third-party runtime dependencies**) that watches Claude Code sessions and triggers shutdown, hibernate, sleep or lock when they are all idle. Built for unattended overnight agent runs.
[](https://www.python.org/) [](#requirements) [](#tests) [](#requirements) [](LICENSE)

The UI is currently in Polish. The screenshot above reads: *Live Claude sessions* — state (WORKING / IDLE), session, directory, **why**, silence, CPU, subagents, PID — and below, *Shutdown conditions (all must be green)*. English localisation is not done yet; PRs welcome.
The problem
You start a long agent run, go to sleep, and want the PC off when it's done. Every naive approach gets this wrong in a way that either **kills work in progress** or **never fires at all**:
| Naive signal | Why it fails |
|---|---|
| "No CPU usage" | An idle Claude Code session and a working one occupy the same CPU range — measured below |
| "Log file hasn't changed in N minutes" | During /compact or an API rate-limit wait the transcript is frozen for minutes while the agent is very much alive |
| "The process exited" | Sessions stay resident between turns; the process living tells you nothing |
| "Window title / focus" | Says nothing about background subagents |
Related Skills
Reel Factory
One raw talking-head recording in, finished vertical video cuts out. Captions, motion-graphic cards, thumbnail
Refine Epics
Break down a MIDAS Epic into sprint-sized User Stories. Reads the Epic content and MIDAS context to generate a
Jev Gates
Three gates for any coding agent: a deterministic approval gate before irreversible actions, a completion gate
Thumbgate Guard
Turn the last agent mistake into a hard prevention rule the agent cannot bypass. Use after a bad tool call, a
Learn Mode
Interactive code tutor — breaks down codebase logic, explains complexities, translates technical concepts for
Trip Publish
The publish surface. Re-publishes an already-public trip site after edits, and lists what is published. Never
Related Agents
Autopilot Implementer
TDD implementer for autopilot's fast/cost-efficient mode. Implements ONE PLAN.md work package test-first, runs
Introspect Dont Guess
Resolve REAL IRIS class / table / column / config names instead of guessing. Use before writing SQL or referen
Brief Agent
Turns a rough design brief into a structured design spec covering screens, flows, components, interaction stat