agentleFS
megandmartinmegandmartin/agent-skills-repo/skills/productivity/inbox-zero-sprint/SKILL.md

inbox-zero-sprint

Skill4 starsChanged 4 months agoMITView on GitHub

Runs a timeboxed 25-minute inbox batch protocol — sort by sender value, apply the 2-minute rule, batch replies as drafts, park and schedule the rest — ending with a hard count report. Use when the user says "inbox zero", "help me clear my inbox", "email sprint", "my inbox is out of control", or wants a focused email session. Don't use for ongoing daily triage rules — use email-triage — or for customer support queues — use customer-support-inbox.

The file

Files
SKILL.md

Inbox Zero Sprint

A 25-minute, clock-driven assault on an overflowing inbox. Not a filing philosophy — a sprint protocol: sort by who's asking, kill the 2-minute items, batch-draft the real replies, park everything that deserves a scheduled slot, and finish with numbers. The rails: every reply is a draft the user sends, nothing is ever deleted, and when the 25 minutes end, the sprint ends — a truthful count beats a fake zero.

When to Use

  • The inbox has piled up (travel, launch week, avoidance) and the user wants one focused session.
  • "Email sprint", "let's clear this out", "inbox zero session".
  • Recurring use is fine — many users run this as a Friday-afternoon ritual — but each run is its own 25-minute box.
  • Not for: the standing daily triage pass — use email-triage. Not for customer support severity queues — use customer-support-inbox. Not a filing/labeling system redesign.

Quick Reference

ActionCommand / Call
Start the clockdate +%H:%M:%S — note it; the sprint ends 25 min later, checked at each phase gate
Load the inboxconnected email tool: list unread/inbox (read + draft only — no send, no delete)
Sender-value tiersT1 clients & money · T2 team & partners · T3 known humans · T4 machines (newsletters, notifications)
Phase budgetSort 5 min · 2-minute kills 8 min · Batch drafts 8 min · Park + schedule 4 min
Park listparked-email.md — thread, why parked, scheduled slot
End mathpython3 -c "print({start_count} - {handled} - {parked})" — must equal what's left
2-minute ruleif a reply takes under 2 minutes to draft, draft it now; the moment it balloons, park it
Follow-up sprintif >15 items remain, propose a second 25-minute slot tomorrow — don't extend today's

Procedure

  1. Precheck — confirm the email tool is connected with draft capability; record the starting unread/inbox count and the start time. Announce the contract: 25 minutes, drafts only, nothing sent, nothing deleted, archive only with the user's OK.
  2. Sort (≤5 min) — sweep the pile into the four sender-value tiers. T4 machines get bulk treatment immediately: newsletters and notifications are marked-read/archived as a batch, with one confirmation ("42 automated messages → archive all? y/n"). Never unsubscribe or delete on the user's behalf.
  3. 2-minute kills (≤8 min) — walk T1→T3 items answerable in under 2 minutes each: one-line confirmations, yes/no answers, quick forwards. Draft each (don't send), mark the thread handled. If an item balloons past 2 minutes, stop mid-thought and move it to the batch pile — the rule is the rule.
  4. Batch drafts (≤8 min) — the substantial replies, highest sender-value first. Full drafts in the user's voice, saved to the drafts folder with the thread. Anything needing research, a decision, or more than ~5 minutes of writing is not drafted now — it's parked.
  5. Park + schedule (≤4 min) — every remaining real item goes to parked-email.md: thread subject, sender, why parked (needs decision / needs research / waiting-on), and a proposed calendar slot. Offer to create the time blocks via the calendar tool — create events only on explicit confirmation.
  6. Count report — stop at 25 minutes even if items remain. Compute the arithmetic, deliver the template, and remind the user their only job now: review the drafts folder and hit send.

Output Template

# Inbox Sprint — {date}, {start}–{end} (25:00)

Start: {N} in inbox
├─ 🗑️ Bulk-archived machines: {n} (with your OK)
├─ ⚡ 2-minute drafts: {n}
├─ ✍️ Batch reply drafts: {n}
├─ 📌 Parked + scheduled: {n} → parked-email.md
└─ 📥 Remaining untouched: {n}
Check: {N} − {archived} − {drafted} − {parked} = {remaining} ✓

**Your move:** {n} drafts waiting in the drafts folder — review and send.
**Parked slots proposed:** {slot} for {item} ...
Nothing was sent or deleted during this sprint.

Pitfalls

  • Sprint dissolves into one 25-minute email — a single T1 thread eats the box. Recovery: the phase gates exist for this; at each date check, if a phase is over budget, park the current item mid-draft and advance. Depth goes in the parked slot, not the sprint.
  • Agent sends "obviously safe" replies — even a one-line "confirmed, see you Tuesday" can be wrong (wrong Tuesday). Recovery: the drafts-only rail has zero exceptions; if a send happened, tell the user immediately and exactly what went out.
  • Bulk archive swallows a human — a real person's mail sat in a newsletter-looking thread and got archived in the T4 sweep. Recovery: before the bulk action, list sender names, not just the count; anything with a personal name gets pulled out of the batch. Archive is recoverable — which is why delete is never used.
  • Fake zero — items quietly archived without drafting or parking to make the number pretty. Recovery: the count arithmetic in the report must balance; any item not drafted, parked, or confirm-archived is "remaining", and that's fine.
  • Parked list becomes a graveyard — parking without a scheduled slot is procrastination with a filename. Recovery: every park entry gets a proposed calendar slot in the same breath; unscheduled parks block the report's "done" line.

Verification

  • Sprint duration honored: end time − start time ≤ ~25 min, phases gated by real date checks
  • Zero sends, zero deletes in the transcript; bulk archive had an explicit user yes with senders listed
  • Count arithmetic balances (python3 check shown in the report)
  • All parked items exist in parked-email.md with a proposed slot each
  • Drafts folder holds every drafted reply, matching the report's draft count
  • If items remain, a follow-up sprint slot was proposed rather than silently overrunning
  • Report delivered even though the inbox isn't at literal zero — truthful count over fake zero

Download now (ZIP)

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
What did you use it for? e.g. Reviewing pull requests in a Rails app

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.