Inbox Triage logo

Inbox Triage

Community
latour-ai
inbox-triage

Pre-market inbox triage for hedge fund executives. Scans Gmail and surfaces only the emails that matter before the trading day starts. Use this skill whenever the user asks to triage their inbox for trading day prep, says "pre-market inbox", "morning email scan", "what's urgent before the open", "check my email", "inbox triage", "morning briefing", "email priorities", "what needs my attention", "any urgent emails", "what did I miss overnight", "catch me up on email", "what's waiting for me", "pre-market check", "morning triage", or any variation of wanting to see what emails need attention before markets open. Also trigger when the user says "trading day prep", "what came in overnight", "anything from investors", "LP emails", "compliance emails", "risk alerts", or mentions needing to scan email before the bell. This skill is purpose-built for hedge fund and investment management executives who receive hundreds of emails daily and need a ruthlessly filtered view of what actually matters.

Overview

Publisherlatour-ai
Repositoryskills
Skill nameinbox-triage
Stars
69
Forks
4
Bundled files
Instructions only
Links
  • Markdown instructions

    A SKILL.md file the model loads on demand, so it only costs tokens when a request actually matches.

  • Works with any LLM

    AI skills are plain Markdown, not provider-specific code, so this works with GPT, Claude, Gemini, Grok, or a local model.

  • Self-contained

    Everything the model needs lives in the instructions — no extra files to sync.

  • Open source

    Published by latour-ai on GitHub. Read the source before you install it.

Installation

Install the Inbox Triage AI skill in TypingMind to use it with any LLM, or drop it into another agent that reads SKILL.md.

1

Install in TypingMind

TypingMind installs a skill straight from its GitHub folder — it reads SKILL.md, bundles the resource files, and stores the result locally.

  1. Open the app and go to Plugins → Skills.
  2. Choose "Install from GitHub".
  3. Paste the skill folder URL below and confirm.
    https://github.com/latour-ai/skills/tree/main/inbox-triage
  4. Enable the skill in any chat where you want it available.
Plugins → Skills → Add skill → From GitHub URL, then paste the folder URL and press Continue.
2

Install in another agent

Any agent that reads the Agent Skills format can use this skill — copy the folder into that agent's skills directory.

Claude Code — .claude/skills
git clone --depth 1 https://github.com/latour-ai/skills.git /tmp/skills
mkdir -p .claude/skills
cp -r /tmp/skills/inbox-triage .claude/skills/inbox-triage
Restart Claude Code after copying so it picks up the new skill.

Use it in TypingMind

Enable Inbox Triage in any TypingMind chat and the model takes it from there. Its name and description sit in the system prompt, and the moment a request matches, the model loads the full instructions itself — you never invoke it by hand, and it costs no tokens until it is actually used.

The model loads Inbox Triage on its own as soon as a request matches it.

Works with any AI model

AI skills are plain Markdown instructions rather than provider-specific code, so Inbox Triage is not tied to the model it was written for. Install it once in TypingMind and use it with GPT-5, Claude, Gemini, Grok, DeepSeek, Mistral, Llama, or a local model you run yourself — all on your own API keys.

  • Loaded only when it is needed

    The system prompt carries just the name and description. The instructions are fetched on the first matching request, so an idle skill costs nothing.

  • Switch models mid-chat

    Because the skill is instructions rather than code, changing model does not break it — the next model reads the same SKILL.md.

Skill instructions

This is the SKILL.md content the model loads. Read it before installing — a skill is instructions your model will follow.

Inbox Triage

Ruthlessly filter a hedge fund executive's inbox down to what matters before markets open. This person gets hundreds of emails a day — the bar for surfacing something is HIGH.

Why this exists

A hedge fund executive's morning is not like a normal morning. Before 9:30 AM ET, they need to know: Is there anything in my inbox that could affect my positions, my investors, my compliance posture, or my team? Everything else is noise. This skill applies that lens mercilessly.

VIP Stakeholders (customize this list)

Messages from anyone on this list are automatically flagged 🔴 HIGH regardless of content. Update these names and emails to match the actual people in your orbit.

Direct Reports To

  • [Your CIO / Managing Partner name]
  • [Fund principal / founder]

Portfolio Managers

  • [PM 1 name]
  • [PM 2 name]

Investors / LPs

  • [Key LP contact 1]
  • [Key LP contact 2]
  • [Placement agent contacts]

Risk & Compliance

  • [Chief Risk Officer]
  • [Chief Compliance Officer]
  • [Risk committee members]
  • [Outside counsel — regulatory matters]

Prime Brokerage & Counterparties

  • [Prime broker contact]
  • [Key counterparty contacts]

Other VIPs

  • [Board members]
  • [Fund administrators]
  • [Key service providers — auditor, legal]

Priority Tiers

🔴 HIGH — Action required before market open

These are emails that could affect positions, capital, compliance, or key relationships. If it's 🔴, the executive should see it before they look at anything else.

Auto-HIGH triggers (regardless of content):

  • Sender is on the VIP Stakeholders list above
  • Someone is following up on something you owe them ("circling back", "any update", "following up", "checking in")

Content-based HIGH triggers:

  • Trade-related communications: execution issues, settlement fails, margin calls, position inquiries, NAV discrepancies
  • Investor/LP communications: capital calls, redemption notices, reporting requests, due diligence questions, investor meeting follow-ups
  • Regulatory and compliance: SEC notices, compliance alerts, regulatory filings deadlines, audit requests, suspicious activity flags
  • Risk alerts: limit breaches, drawdown notifications, counterparty credit issues, liquidity concerns
  • Urgent operational: fund administrator issues, prime brokerage communications, wire transfer confirmations/issues, legal matters with deadlines
  • Client escalations: unhappy investors, service complaints, performance questions with urgent tone
  • Time-sensitive language: "before the open", "EOD", "ASAP", "urgent", "immediately", "deadline today", "time-sensitive", "waiting on you", "need this by"

🟡 MEDIUM — Important but can wait until mid-day

These matter, but the fund won't blow up if they wait a few hours.

Content-based MEDIUM triggers:

  • Meeting prep for today's calendar (board meetings, investor calls, PM reviews) — especially if there are materials to review
  • Non-urgent colleague requests: research asks, process questions, team coordination
  • Counterparty communications without urgency: documentation requests, standard confirms, relationship check-ins
  • Operational items with deadlines this week but not today
  • New business / fundraising threads that aren't time-critical
  • Warm introductions and referrals — these are relationship-valuable but can wait
  • Legal documents for review (not deadline-imminent)
  • HR and compensation matters

🟢 LOW — Worth knowing, no action needed today

Good to be aware of, but zero urgency.

Content-based LOW triggers:

  • Industry news that's genuinely relevant to the fund's strategy (not generic market commentary)
  • Conference and event invitations
  • Completed document signatures (DocuSign, etc.) — confirmation only
  • Internal announcements that are informational
  • Professional network updates that are actually substantive
  • Building/office logistics (visitor passes, facility notices)

What to IGNORE entirely

These never appear in the output. Don't even count them in the filtered total — they're invisible.

Newsletters & digests:

  • Any sender containing "substack", "newsletter", "digest", "noreply", "no-reply", "mailer", "notify", "updates@", "news@"
  • Domains: beehiiv.com, substack.com, mailchimp.com, hubspot.com, intercom.io
  • Market commentary blasts, research distribution lists, morning notes from sell-side
  • AI newsletters (The Rundown, Ben's Bites, TLDR, etc.)
  • Pebblr daily digests (noreply@pebblr.io)
  • Substack notification emails and own-post delivery copies

Research & market blasts:

  • Sell-side research distribution (Goldman, JPM, Morgan Stanley morning notes, etc.)
  • Market data service alerts (Bloomberg, Refinitiv, FactSet automated emails)
  • Generic market commentary and "daily wrap" emails from brokers
  • Bulk research reports — unless from a PM or analyst on the VIP list forwarding something specific

Promotional & transactional:

  • SaaS product emails, billing receipts, welcome emails
  • Vendor pitches, demo requests, cold sales outreach
  • Google Workspace notifications, product updates
  • Subscription confirmations, order receipts

Calendar noise:

  • "Accepted:", "Declined:", "Tentative:" auto-responses (unless they contain a personal note)
  • Google Calendar notification sender (calendar-notification@google.com)

Low-signal social:

  • LinkedIn notifications, Twitter/X notifications
  • Social media digests
  • Group threads that are clearly social banter (short messages, jokes, emojis)
  • "Thread closed" replies — short acknowledgments like "Thanks!", "Got it", "Sounds good!" that signal a conversation is resolved

Automated system alerts:

  • CI/CD notifications, deployment alerts
  • IT system monitoring (unless it's a trading system outage — use judgment)
  • Password reset emails, 2FA codes (these are ephemeral)

Workflow

Step 1 — Fetch recent inbox messages

Run two targeted Gmail searches. Default time window: overnight (last 12 hours). Adjust based on what the user asks — "what did I miss this week" = 7 days, "anything urgent" = last 6 hours.

Search 1 — Direct emails:

Query: "is:unread -category:promotions -category:social newer_than:1d"
maxResults: 75

Search 2 — CC'd emails (shorter leash):

Query: "is:unread cc:me -category:promotions -category:social newer_than:1d"
maxResults: 30

Use the Gmail MCP tools (prefixed mcp__6c627729-09a2-40da-97c1-850fb310e12e__) for all email operations:

  • search_threads to run the queries above
  • get_thread to read full message content for emails that survive first-pass filtering

Deduplicate results by thread ID across both searches. Track which emails had the user in the To field vs. CC — To-field emails get priority over CC in borderline cases.

Step 2 — First-pass filtering (metadata only)

Using only sender, subject line, and snippet from search results, immediately discard anything matching the IGNORE patterns above. No need to read full messages for obvious noise.

Be aggressive here. The goal is to cut the list from potentially 75+ emails down to the 10-20 that deserve a closer look.

Step 3 — Read and classify surviving emails

For each email that survived the first pass, use get_thread to read the full content. Apply the priority tier rules:

  1. Check VIP list first. If the sender matches anyone in the VIP Stakeholders section, it's automatically 🔴 HIGH.
  2. Check for "waiting on me" signals. Follow-ups, unanswered questions, threads where the user was last asked something — these are 🔴 HIGH regardless of sender.
  3. Apply content-based classification using the tier definitions above.
  4. CC vs. To tiebreaker. If an email is borderline between two tiers and the user is only CC'd (not in the To field), drop it to the lower tier.

If an email still looks like noise after reading the full content, drop it silently.

Step 4 — Present the briefing

Header line with counts:

📬 X emails scanned, Y surfaced (Z high / W medium / V low), N filtered out

Then organize by priority tier:

🔴 HIGH — Before the open
**[Sender Name]** — [Subject line]
[One-line summary: what happened / what they need]
⏩ Action: [Reply / Forward to [person] / Review and decide / Schedule call / Escalate]
🕐 Received: [time, e.g., "6:42 AM" or "Yesterday 11:15 PM"]
🟡 MEDIUM — Mid-day
**[Sender Name]** — [Subject line]
[One-line summary]
⏩ Action: [Reply / Review / Schedule / FYI only]
🕐 Received: [time]
🟢 LOW — When you have a moment
**[Sender Name]** — [Subject line]
[One-line summary]
⏩ Action: [FYI only / Review later / No action needed]
🕐 Received: [time]

Footer:

Filtered out: ~N newsletters, ~N research blasts, ~N promotional, ~N calendar/system noise

Important Behaviors

  • The bar is HIGH. This is not a general inbox summary. If you're unsure whether something deserves to surface, lean toward filtering it out for 🔴 and 🟡, but keep borderline items as 🟢 rather than dropping them entirely. It's worse to miss a genuine issue than to include one extra low-priority item.
  • Direct-to beats CC. If the user is in the To field, the email is more important than if they're CC'd. CC-only emails need strong signals (VIP sender, urgent language, trade-related content) to reach 🔴.
  • Context matters for compliance and risk. An email from "compliance@" that says "annual training reminder" is noise. An email from compliance that says "urgent — position limit breach" is 🔴. Read before classifying.
  • Overnight vs. intraday. If the user runs this before market open, weight anything that arrived overnight or early morning more heavily — those are things they haven't seen yet. If they run it mid-day, adjust accordingly.
  • Empty inbox is a good sign. If there's nothing urgent, say so: "Your inbox is clean — nothing requiring attention before the open." Don't manufacture urgency.
  • Adapt to specificity. If the user asks about a specific sender, topic, or timeframe ("anything from our prime broker?", "LP emails this week"), narrow the search instead of running the full triage.
  • Respect the clock. Keep the output tight and scannable. This person has 15 minutes before the open, not 15 minutes to read your email summary.

Frequently asked questions

What does the Inbox Triage AI skill do?

Pre-market inbox triage for hedge fund executives. Scans Gmail and surfaces only the emails that matter before the trading day starts. Use this skill whenever the user asks to triage their inbox for trading day prep, says "pre-market inbox", "morning email scan", "what's urgent before the open", "check my email", "inbox triage", "morning briefing", "email priorities", "what needs my attention", "any urgent emails", "what did I miss overnight", "catch me up on email", "what's waiting for me", "pre-market check", "morning triage", or any variation of wanting to see what emails need attention...

Why use Inbox Triage on TypingMind?

Because you install it once and use it with any model. Inbox Triage is plain Markdown rather than provider-specific code, so the same skill runs on GPT-5, Claude, Gemini, Grok, or a local model — and you can switch model mid-chat without it breaking. TypingMind runs on your own API keys, so you pay providers directly instead of a per-seat subscription, and your skills and chats stay in your own storage.

How do I install Inbox Triage in TypingMind?

Open Plugins → Skills → Install from GitHub in TypingMind and paste https://github.com/latour-ai/skills/tree/main/inbox-triage. TypingMind reads its SKILL.md and installs it as a skill you can enable per chat.

Which AI models can use Inbox Triage?

Any model you connect in TypingMind. AI skills are plain Markdown instructions rather than provider-specific code, so GPT, Claude, Gemini, Grok, and local models can all load this skill when a request matches it.

How many AI models can I use with Inbox Triage?

As many as you like. As long as a model supports skills, you can use Inbox Triage with it — GPT, Claude, Gemini, Grok, DeepSeek, Mistral, Llama and more — all on TypingMind with your own API keys.

Is the Inbox Triage AI skill free?

It is published on GitHub by latour-ai. Check the repository for licensing terms. You only pay your own AI provider for the tokens you use.

What are AI skills?

An AI skill is a reusable instruction bundle that teaches an AI model how to do one specific task. It follows the open Agent Skills format: a SKILL.md file with a name and description, plus any scripts, templates or reference files the model may need. The model reads the instructions only when your request matches the skill, so an installed skill costs nothing until it is used.

How are AI skills different from plugins or MCP servers?

A plugin or MCP server gives a model new tools to call — code that runs somewhere and returns a result. An AI skill gives the model knowledge and process instead: how to approach a task, which steps to follow, what good output looks like. Skills are plain Markdown, so they need no server, no API key and no runtime, and they work with any model.

View all

Set up your own AI workspace now

Get notified about new features and future giveaways by subscribing to our newsletter 👇