agentsclimarketplace

Inbox triage

Skill TheCraigHewitt/cowork-starter-pack/skills/inbox-triage

Processes a batch of emails — categorizes them, drafts responses where appropriate, and tells the user which ones need a human decision. Use when the user asks to triage the inbox, clear email, batch-process email, or prep email responses. Best run once or twice a day, not continuously.From its SKILL.md

Install
npx -y skills add TheCraigHewitt/cowork-starter-pack --skill inbox-triage

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

  • 2 stars2 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

SKILL.md

5.1 KB, ~1.2k tokens by cl100k_base, as published. Nobody here has run it

Inbox Triage

Turn a full inbox into three stacks: handled, drafts-for-your-review, and flagged-for-a-decision. Don't send anything without explicit approval.

Before starting

  1. Read context/about-me.md — who is the user, who are their priority contacts
  2. Read context/voice.md — any draft reply must sound like them
  3. Read context/preferences.md — their rules on what to auto-reply vs. escalate
  4. Read memory.md if it exists — prior corrections to triage logic

What to produce

A markdown file saved to /output/inbox-triage-{YYYY-MM-DD-HHmm}.md with four sections:

  1. Archived / filed (no action needed) — count + one-line summary. The newsletters, receipts, notifications, FYI-cc'd threads. Just counts, not individual items.
  2. Drafted replies for your review — threads where a standard response is appropriate and you've drafted it. Each item: sender, subject, one-line context, draft reply.
  3. Flagged for your decision — threads that need the user's judgment before anyone replies. Each item: sender, subject, one-line context, the specific decision needed.
  4. Summary — total emails processed, how many in each bucket, any patterns worth noting.

Do not actually send, archive, or label anything unless the user has explicitly approved auto-actions in their preferences file.

Inputs

  • Gmail or equivalent connector (required)
  • The window to triage (default: since last triage run, or last 12 hours)
  • Priority contacts list from about-me.md or preferences.md

How it works

  1. Pull emails from the window (unread + recently-read)
  2. For each email, categorize:
    • Archive-worthy? Newsletters, receipts, automated notifications, CC threads where user isn't the primary recipient, confirmations that don't require action
    • Draftable? Clear request with an obvious standard answer — meeting confirmations, simple scheduling, "thanks" replies, known-answer FAQs
    • Needs decision? Everything else that isn't obvious
  3. For draftable items, write a reply in the user's voice. Under 80 words unless the situation needs more.
  4. For decision items, name the specific decision needed ("accept/decline," "yes/no/renegotiate," etc.)
  5. Save the triage file, tell the user what to look at first

What not to do

  • Do not send anything without approval. This is a triage, not an autoresponder.
  • Do not draft replies for anything that involves: money, a commitment, a promise, a new relationship, or legal language. Flag those.
  • Do not apply labels, move to folders, or archive without explicit prior permission in preferences.md.
  • Do not hallucinate context from email threads. If a thread is ambiguous, flag it — don't guess.

Example output structure

# Inbox Triage — 2026-04-18 10:42 AM

## Summary
- 47 emails processed since 2026-04-18 07:00
- 31 archive-worthy (newsletters, receipts, notifications)
- 9 drafted replies for your review
- 7 flagged for your decision

**Look at first:** Sarah Chen's reply about the SOW — she's asking a pricing question I can't answer.

---

## Archived / filed (31, no action needed)
- 12 newsletters
- 8 automated notifications (calendar, Slack digest, GitHub)
- 6 CC'd threads where you weren't the primary recipient
- 5 receipts / order confirmations

## Drafted replies for your review (9)

### 1. Morgan — "Tuesday 1:1 — reschedule?"
Context: Morgan needs to move the 1:1 from 9am to 11am Tuesday
Draft reply:
> Morgan — 11am works. Same agenda. See you then.

### 2. Conference organizer — "Confirming your speaker slot"
Context: Routine confirmation for a talk you already agreed to
Draft reply:
> Confirmed. I'll send the slide deck by 2026-04-26 as requested. Let me know if the tech check window changes.

[...7 more...]

## Flagged for your decision (7)

### 1. Sarah Chen (Northwind) — "Quick question on pricing"
Context: Sarah is asking for a 15% discount in exchange for a multi-year commitment. This is a pricing decision I can't make for you.
Decision needed: accept / counter / decline with alternative

### 2. Taylor (vendor) — "Contract renewal — new terms attached"
Context: New contract has a 12% price increase and changed liability terms.
Decision needed: review and send back, or route to legal

[...5 more...]

Customization notes

  • For EAs. You're likely triaging both your inbox and a principal's inbox. Ask which inbox, and tune the "priority contacts" list accordingly.
  • For high-volume inboxes. Set the triage window to 4–6 hours instead of 12. Run it more often.
  • For founders. Be more conservative on "drafted replies" — when in doubt, flag for decision. The risk of a wrong-voice email going out is higher than the time saved.
  • If the drafts are bad. Save feedback to memory.md. The next run will be better. This skill improves fast with correction.

What ships with it

Read from the repository

Just SKILL.md. No reference files, no scripts.

Gives 0 of the 12 instructions most debug triage skills give in ~1.2k tokens

Counted across 1,020 of the 1,639 authors here whose files we hold, read 2026-09-06

  • Find root cause before attempting any fixin 134 of 1020, across 118 files
  • Create a failing test case before implementing a fixin 109 of 1020, across 95 files
  • Read error messages and stack traces completelyin 102 of 1020, across 88 files
  • Reproduce the issue consistently before investigatingin 90 of 1020, across 77 files
  • Make the smallest possible change to test a hypothesisin 90 of 1020, across 76 files
  • Trace data flow backward to find the sourcein 84 of 1020, across 70 files
  • Form a single hypothesis before testingin 78 of 1020, across 64 files
  • Implement only one fix at a timein 76 of 1020, across 63 files
  • Question the architecture if three fixes failin 73 of 1020, across 59 files
  • Add diagnostic instrumentation at component boundariesin 68 of 1020, across 56 files
  • Compare broken code against working examplesin 68 of 1020, across 57 files
  • Write a regression test before applying the fixin 62 of 1020, across 55 files

Said here and by no other author read

  • categorize emails into three stacks
  • flag ambiguous threads for user decision
  • save triage results to a markdown file
  • notify the user of the triage completion

Grouped from the skills themselves: near-identical wordings counted once, and counted by distinct author, so one author publishing three of these counts once. Length counted with cl100k_base; the agent that loads this file may tokenize it differently.

Keep looking

Skills are one crate of 325,949. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.