Loopany
Sign in
Back to templates
Codebase Autopilot

Doc Maintainer

A weekly sweep that catches documentation drift, verifies what really works, and lands fixes as a PR.

Moderate setupLong cycleOpen loopBuilds over runs
Setup in Loopany

How a run flows

On schedule
Every Mon · 6am
Step 1 · Compare
Diff since last sweep
Step 2 · Verify
Verify for real
Isolated git worktree · off main
Step 3 · Fix
Fix real drift only
Step 4 · Ship
Open a PR
Report
Latest sweep summary
Metric
Drift count
open loop — repeats on the next tick, picking up where it left off

Prerequisites

How this loop works — schedule, exit, notifications
Suggested schedule
Weekly · Mon ~6am
Loop type
Open loop — an ongoing monitor with no finish line.
Exit condition
Runs indefinitely (no finish line).
Notifications
On by default — you're alerted on each run's outcome (completion or failure). Silence with notify: never.
Ease of starting
Some setup and context are needed before the first useful run.
When you see value
Each caught drift lands as a PR; a well-kept codebase may take a few sweeps to surface one.

This month, Docs Sweep caught two stale claims. Our contributing guide still named a database driver we'd stopped using, and it described a migration flow we'd already replaced. Either one could send a new contributor down a pretty long detour.

On Mondays, the loop checks the claims it can actually prove. Commands get run. It checks flags and paths against the repo, along with environment variables and versions. Opinions and tone aren't part of the job.

The loop's dashboard: every Monday accounted for, and the latest sweep showing what the docs say against what is true now

When the docs have drifted, it opens a small PR showing what they say and what's true now. We've shipped 6 of those in the last two months. Median review was under a minute. Good docs still rot. The weekly check keeps them honest.

The prompt this runs

Set up a weekly loop that keeps this project's documentation honest. Every Monday around 6am my local time, review what has changed since the previous sweep - the loop keeps track of where the last one ended - and compare the README, setup guides, examples, and runbooks against what the code actually ships now. Verify for real: run the commands, check the links, try the examples rather than trusting memory. Fix only material that is genuinely stale - never rewrite accurate docs just to create activity; finding nothing wrong is a clean stop. Do the work in a fresh git worktree off the main branch and located outside the loop folder - the loop folder is a synced content home, not a workspace - so neither my working checkout nor the sync is polluted, and open a pull request explaining the drift and the fixes - but if a previous sweep's PR is still unmerged, don't stack another one. Write a short summary each sweep and report the drift count as a metric, zero included. Set up the loop's dashboard at creation: an embed of the latest sweep summary plus a chart of the drift count over time.

This is the exact intent appended to the bootstrap when you create the loop — your coding agent reads it, then proposes cadence and config and confirms with you before creating anything.