runbookify
← All plans
Project & Work Management / Task & Milestone Tracking

Stale Task Sweeper & Owner Nudger

Build an internal tool that scans your task tracker for items going quiet, groups the stale ones by owner, and drafts a gentle nudge for a status or a new date — with a PM approval gate before any email goes out.

BeginnerAn afternoonBuilds onNext.js (App Router) on VercelSupabase (Postgres, Storage, Auth + RLS)Resend (email)
What you'll build

A private web tool where you import tasks, see exactly which ones have gone stale, review them grouped by owner, approve the nudge batch, and send polite status-request emails — with every nudge logged and a clean CSV export.

Gated download

Enter your email — the plan downloads instantly and a copy lands in your inbox.

By submitting your email you'll also receive the weekly runbookify newsletter. You can unsubscribe at any time.

Before you start

  • A task list with owner, status, last-updated, and due date (CSV or Google Sheet export)
  • A people list (name + email) for your owners
  • Free accounts: Vercel, Supabase, Resend

The problem this kills

You manage the work, so you become the human cron job. Every few days you scroll the tracker hunting for tasks that have gone quiet — no update, no comment, due date creeping closer — and then you hand-write the same "hey, where are we on this?" message to a dozen different people. It's tedious, it's easy to miss things, and half the items you nudge were already handled (you just annoyed someone for nothing).

The work isn't hard. The chasing is. And chasing is exactly the kind of repetitive, rules-based job that an internal tool should do for you — drafting the nudges while you stay the human who decides what actually goes out.

What you'll build

A small private web app, just for your team, that:

  • Imports your task list from a CSV or Google Sheet (owner, status, last-updated, due date).
  • Finds the stale items — anything with no update in N days and no recent comment — while ignoring done and cancelled work.
  • Groups the stale tasks by owner so you see "Priya has 4 quiet items, Marcus has 2" at a glance.
  • Drafts a gentle nudge per owner asking for a status or a fresh date — and flags items old enough to escalate.
  • Waits for you. Nothing sends until you, the PM, review the stale list and approve the batch.
  • Sends and logs. Approved nudges go out by email, every nudge is recorded, and a duplicate guard stops the same task from being nudged again inside your cooldown window.
  • Exports a clean CSV of the stale list and what was sent, in the columns your system expects.

It's realistic for one afternoon of building, even if you've never written code.

What's inside the Implementation Plan

A complete, paste-and-go runbook for an AI coding agent (Claude Code). The most important part: it opens by interviewing you about your business — your tracker, your status names, how "stale" actually works for your team, your escalation rules, your edge cases — and then tailors the whole tool to your answers instead of dropping a generic template on you. You confirm a short spec, then it builds.

Inside you get:

  • The discovery interview (8–12 tailored questions) and a spec to confirm.
  • The exact data model for tasks, people, nudges, and the audit log.
  • Step-by-step build prompts you copy and paste — import, the staleness engine, the owner grouping, the draft-and-review screen, the approval gate, sending via Resend, logging, and the CSV export.
  • A configurable staleness threshold per status, plus a cooldown window so nobody gets nudged twice.
  • The "No API yet?" fallback so you can build the entire thing today from a spreadsheet, no integration required.
  • A verification checklist so you know it actually works.

The governance it includes (this is the point)

This isn't a script that blasts emails. It's an internal tool built like a grown-up system:

  • Login so only your team can open it.
  • Row-level security so each organization only ever sees its own tasks and people.
  • A complete audit trail — who approved which nudge, to whom, and when.
  • A hard human-in-the-loop approval gate — the AI drafts, you review and approve, and only then does anything send.
  • Duplicate guards — a cooldown window so the same task can't be nudged twice, and so a re-import can't double-send.

Who it's for

Project managers, team leads, and PMO coordinators who are tired of being the one who chases. If you live in a task tracker or a project spreadsheet and you're the person who has to ask "any update?" all day, this turns that grind into a five-minute review-and-approve.

You've got this — paste the first prompt and let the interview tailor it to your team.

Gated download

Enter your email — the plan downloads instantly and a copy lands in your inbox.

By submitting your email you'll also receive the weekly runbookify newsletter. You can unsubscribe at any time.