agentleFS
TheCraigHewittTheCraigHewitt/cowork-starter-pack/skills/inbox-triage/SKILL.md

inbox-triage

Skill5 starsChanged 6 months agoMITView on GitHub

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.

We split this skill into the rules it gives an agent

9 rules. 0 are common, 4 are distinctive, and 5 are very specific to this project.

  • 0

    Common

    Most of the 42 edited versions we read keep these.

  • 4

    Distinctive

    This author's own choices. Worth a look.

  • 5

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Read context/about-me.md — who is the user, who are their priority contacts

    line 12
  2. Very specific:

    Read context/voice.md — any draft reply must sound like them

    line 13
  3. Very specific:

    Read context/preferences.md — their rules on what to auto-reply vs.

    line 14
  4. Very specific:

    Read memory.md if it exists — prior corrections to triage logic

    line 15
  5. Distinctive:

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

    line 26
  6. Distinctive:

    Do not send anything without approval.

    line 47
  7. Distinctive:

    Do not draft replies for anything that involves: money, a commitment, a promise, a new relationship, or legal language.

    line 48
  8. Very specific:

    Do not apply labels, move to folders, or archive without explicit prior permission in preferences.md.

    line 49
  9. Distinctive:

    Do not hallucinate context from email threads.

    line 50

“Common” is measured: kept by at least 6 in 10 of the 42 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.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

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.

Download now (ZIP)

  • Other versions49 projects use a version of this.

    Word for word · 0

    None found.

    Edited · 0

    None found.

    Wrote their own version · 49

    1. gnekt/My-Brain-Is-Full-Crew★ 3.9ksame name
    2. aeonfun/aeon★ 785same name
    3. aaronjmars/aeon★ 776same name
    4. QuixiAI/agi-memory★ 617same name
    5. QuixiAI/Hexis★ 617same name
    6. microsoft/cat-agent-skills★ 95same name
    7. latour-ai/skills★ 69same name
    8. kesslernity/awesome-copilot-cowork-skills★ 40same name
    9. GTMify/aigtm★ 25same name
    10. aaronjmars/miroshark-aeon★ 19same name

    39 more not listed

    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.

  • Used alongsideProjects with it also use branded-deck-builder, cmo-reviewer and 10 more.

    Other files people use alongside this one.

    Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
    • kesslernitySkill

      branded-deck-builder

      Builds PowerPoint decks on the user's own company template and checks every slide against written brand rules (colours with hex codes, fonts, logo placement, banned visuals). Use when the user asks for a branded, on-brand, CI-compliant or corporate-template presentation, or wants a deck that follows company design guidelines. Confirms an outline first, then builds slide by slide and saves a DRAFT deck to OneDrive.

      Also in 4 projects that use this

    • kesslernitySkill

      cmo-reviewer

      Reviews a proposal, business case, deck or plan in character as a Chief Marketing Officer archetype, producing a structured DRAFT review document with a verdict, findings cited to specific passages, risks and five interrogation questions. Use when the user asks for a CMO review, a marketing executive's read on a document, or wants to pressure-test positioning, messaging, pricing or a launch plan before a real executive review.

      Also in 4 projects that use this

    • aaronjmarsSkill

      aeon-update

      Pull framework updates from the upstream Aeon repo into this instance - 3-way merges canon's new commits into a PR, never clobbering operator config.

      Also in 4 projects that use this

    • +9 more from kesslernity
  • In this repository6 other files it gives its agents.

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.