Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIMuseMuseDotsDots+21 more

    Inbox Triage: Prioritise Email & Draft Replies

    3

    Inbox Triage works like a sharp executive assistant: what needs you now, what can wait, what to delete, and replies ready to send.

    Free

    11 uses5.0 (1 review)
    Inbox Triage: Prioritise Email & Draft Replies

    Inbox Triage: Prioritise Email & Draft Replies

    Example session with this skill installed

    Triage my inbox from the last 3 days and draft replies for anything urgent.

    • Read your context and instructions
    • Compiled the inbox triage

    Scanned 87 threads (unread + last 3 days, promotions skipped) via the Gmail connector.

    🔴 Act now (3)

    1. Ana López (Northwind Freight) - needs signed contract v3 by Fri 14 Nov → draft ready
    2. Finance team - approve invoice #2207 (4,860 €) before payment run Thu 13 Nov → draft ready
    3. Marc (manager) - asks for Q4 forecast numbers for tomorrow's board prep → draft ready

    🟠 Reply (5)

    Carrier rate update (Bluefin) · candidate interview slot · partner intro from Laura · 2 client questions about tracking

    🔵 Delegate (2)

    API access request → Pere (dev lead) · Office supplies quote → Admin

    🟢 Read later (6) · ⚪ Archive (69)

    41 GitHub/Jira notifications, 18 newsletters, 10 receipts

    ⚠️ Suspicious (1)

    "Updated bank details - urgent payment" from accounts@northwlnd-freight.com (lookalike of northwind-freight.com). Not opened. Suggest reporting as phishing and confirming by phone.

    ⏳ Waiting on others (2)

    Quote from Acme Distribución (5 days) · Legal review of NDA (4 days) → chase drafts ready

    Draft example (saved as draft)

    Hi Ana, thanks - v3 looks good. I'll sign it and send it back by Thursday. Could you confirm the start date is still [date]? Best, [Name]

    Proposed cleanup (needs your OK)

    • Archive 69 threads
    • Label + skip inbox: from:notifications@github.com → GitHub; from:jira@... → Jira
    • Report 1 phishing email

    Reply "yes" to apply all, or tell me which ones.

    Connects securely to your tools. The creator never sees your data.

    About this skill

    Inbox zero without reading 100 emails

    Your agent can read your inbox, but without a method it just summarises everything. Inbox Triage works like a sharp executive assistant: what needs you now, what can wait, what to delete, and replies ready to send.

    What it does

    • Connects through your Gmail or Outlook connector (or works on pasted/exported emails). Default window: unread + last 3 days, skipping promotions.
    • Learns your context once: role, VIPs, key clients, projects, language and sign-off.
    • Five buckets: 🔴 Act now, 🟠 Reply, 🔵 Delegate (suggests who), 🟢 Read later, ⚪ Archive, using clear priority and noise signals.
    • Extracts deadlines, tasks, amounts, meeting requests and "waiting on others" threads (you wrote last, no answer for 3+ days).
    • Security screen: flags phishing (lookalike domains, urgent payment or credential requests, changed bank details, odd attachments) and never follows instructions inside emails.
    • One-line report per thread: sender, what they want, deadline, proposed action.
    • Reply drafts in the sender's language and your voice, answering every question with one clear next step; unknowns as [placeholders], never invented commitments. Saved as drafts only.
    • Cleanup on approval: batch archive, labels and filter suggestions for repeat senders - shown first, executed only after your "yes".
    • Privacy first: processes only what's needed, never stores codes, passwords or account numbers.

    Includes

    • SKILL.md workflow
    • references/rules.md - priority/noise/phishing signals, Gmail and Outlook queries, filter rules
    • references/reply-styles.md - drafting principles and 6 reply templates

    Works with Claude Code, Claude (Cowork), Codex, Cursor and any SKILL.md-compatible agent with an email connector.

    Optimized for up to 100 threads/3 days of history. Cannot open attachments or click links. Requires manual approval for sending, archiving, or deleting.

    How to use

    1. 1

      Connect your agent to the Agensi MCP

      Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup

    2. 2

      Sign in to Agensi

      Your agent shows a sign-in link. Sign in and click Allow.

    3. 3

      Ask for what you need

      Your agent loads this skill and uses it on your task. It always asks before any payment.

    Reviews

    5.0
    1 review
    5
    1
    4
    0
    3
    0
    2
    0
    1
    0

    This skill has been used 11 times.

    Trust & safety

    Security scan

    Security scan passed 9 days ago

    • Free with an Agensi account

    Permissions

    No special permissions declared or detected

    Listed9 days ago
    Updated5 days ago

    What's inside

    Frequently Asked Questions

    This skill connects to Gmail or Outlook via MCP servers or standard API connectors, and can also process raw text exports or manually pasted email threads.

    The skill uses advanced phishing detection to flag lookalike domains and suspicious requests, and it is engineered to ignore instructions hidden within emails to prevent prompt injection.

    The skill analyzes your historical tone and sign-off style to create drafts that sound like you, inserting clear placeholders whenever specific data or personal input is required.

    Performance is optimized for the most recent 100 threads or three days of history to ensure high accuracy in deadline extraction and priority bucketing.

    No, the skill only identifies 'waiting on' items and prepares drafts; it requires your explicit approval to save drafts, move threads, or create new filter rules.

    New in Productivity