overnight
Skill4 starsChanged 5 months agoMITView on GitHub
Schedule the current task to run unattended overnight (default 2 AM) via headless `claude -p`, isolated on a git branch, with results auto-surfaced in your next morning session. Use when the user says "do this at 2am", "run this while I sleep", or "queue this for tonight".
8 other projects use a version of this.
Adopted by
Most recently taken up by
- Celorus/celorus-plugintheir own version, first seen Nov 6, 2026
- Kooooojun/skillstheir own version, first seen Oct 25, 2026
- emanueledenaro/claude-skillstheir own version, first seen Oct 15, 2026
- drewmerc302/claude-skillstheir own version, first seen Oct 13, 2026
- Cleveland-RD/hivehousetheir own version, first seen Oct 13, 2026
- jdurey/skillstheir own version, first seen Jul 19, 2026
Other versions8 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 8
- Cleveland-RD/hivehouse★ 46same name
- Kooooojun/skills★ 1same name
- Celorus/celorus-plugin★ 0same name
- choism4/overnight-skill★ 0same name
- drewmerc302/claude-skills★ 0same name
- emanueledenaro/claude-skills★ 0same name
- Evil-Bane/argus★ 0same name
- jdurey/skills★ 0same name
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository18 other files it gives its agents.
- biz-validateSkill
- braindumpSkill
- catch-upSkill
- clone-websiteSkill
- double-checkSkill
- explain-for-grandmaSkill
- heavy-buildSkill
- heavy-reviewSkill
- heavySkill
- make-promptSkill
- monitor-claudeSkill
- orientSkill
- reelSkill
- sanity-checkSkill
- save-sessionSkill
- shipSkill
- what-breaksSkill
- yt-tranSkill
The file
We split this skill into the rules it gives an agent
1 rule. 0 are common, 1 is distinctive, and 0 are very specific to this project.
0
Common
Most of the 7 edited versions we read keep these.
1
Distinctive
This author's own choices. Worth a look.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 27
MaxTurns: default 60; raise for large build tasks, lower (1–3) for trivial ones.
“Common” is measured: kept by at least 6 in 10 of the 7 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Overnight runner
Captures the task being discussed right now, schedules a Windows Task
Scheduler job that runs claude -p headless at the chosen time (the machine
wakes itself), and writes a result that the morning session shows automatically.
Files: scripts live in ~/.claude/overnight/bin/. You orchestrate; the
PowerShell does the work.
Steps
1. Settle the parameters
Infer from the conversation; ask only if genuinely ambiguous.
- Schedule →
RunAt(localyyyy-MM-dd HH:mm). Default next 02:00 (tomorrow if 2 AM already passed today). Parse "in 90m", "tonight", "at 3", "tomorrow 04:00" relative to the current date/time in context. Must be in the future. - Mode:
buildif the task writes code/files;researchif it only investigates/reads/searches and produces a written report. - ProjectDir: the current working directory (absolute path).
- MaxTurns: default 60; raise for large build tasks, lower (1–3) for trivial ones.
- Goal: one crisp human-readable line (shown in the morning summary).
2. Write a self-contained brief
The overnight run starts fresh — it sees only this brief plus the project
files, NOT this conversation. So the brief must stand alone. Write it to a temp
file (e.g. ~/.claude/overnight/_brief_tmp.md) containing:
- Objective — what to accomplish.
- Context — everything relevant from this conversation: decisions made, constraints, paths, prior findings, links.
- Deliverables — concrete outputs (files to create/modify, or the report to produce).
- Definition of done — how the run knows it succeeded.
- Guardrails — what NOT to touch. For build tasks, instruct it to leave changes in the working tree and NOT run git commit/push — the runner commits to an isolated branch automatically. Telling it to commit wastes turns and can hit the turn limit.
Be generous and explicit — under-specifying is the main failure mode.
3. Schedule it
Run (Windows, pwsh):
pwsh -NoProfile -File "$HOME/.claude/overnight/bin/new-job.ps1" `
-Goal "<one line>" `
-BriefPath "$HOME/.claude/overnight/_brief_tmp.md" `
-RunAt "2026-01-01 02:00" `
-Mode build `
-ProjectDir "<cwd>" `
-MaxTurns 60Optional: -Model opus, -AllowedTools "Read Edit Write Glob Grep Bash",
-NoWorktree (build directly in the project with no branch isolation — only if
the user insists).
4. Confirm
Relay the script's confirmation block (id, fire time, mode, branch, wake status).
Then delete the temp brief. Tell the user: results appear automatically at the
start of their next session; build work lands on branch overnight/<id> (review
with git checkout), main untouched.
Reliability notes (mention only if asked)
- Machine wakes via Task Scheduler
-WakeToRun; missed starts run on next wake (-StartWhenAvailable). - One retry on failure, then a lighter-model fallback on overload; a
result.mdis always written, even on crash. - Runs serialized (global mutex) so two jobs never contend.
- Cancel:
Unregister-ScheduledTask 'OvernightClaude-<id>'. List:Get-ScheduledTask 'OvernightClaude-*'. - On subscription plans, headless
claude -pmay draw from a separate Agent SDK credit pool — check your plan's current terms before relying on heavy unattended runs.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

