Inbox Pro

What this skill does

• Triages the inbox into clear categories and decides one action per email (Inbox Zero)
• Applies the 2-minute rule & one-touch: short things straight to a draft, longer ones to a task
• Pulls out tasks, deadlines, dates and waiting-for (GTD) — exportable / to the Weekly Planner
• Drafts replies in the right tone (default concise & friendly, adjustable)
• Summarizes long threads (who wants what, open points, next step)
• Detects phishing/social engineering and flags it instead of acting
• Only drafts — never sends, deletes, forwards or changes subscriptions/filters without approval
• Invents nothing: placeholders instead of made-up promises; CC = informed, not asked
• Suggests newsletters to unsubscribe/filter and recommends batch processing
• Handles appointment/calendar invites: accept/decline/reschedule draft + calendar-entry suggestion
• Replies in the language of the incoming email (DACH/international) and matches formality to the relationship
• De-escalates angry senders and phrases polite declines instead of a hard no
• Creates reusable templates (canned responses) for recurring mail types
• Doesn't open attachments/links unchecked and warns on suspicion
• Classifies mail types (personal, distribution/CC, newsletter, system notification, transactional, offer, request, spam …) and derives the action
• Phishing heuristics with typical signs (lookalike domain, display name ≠ address, pressure, bank-detail change, link mismatch, CEO fraud); with web access, optional domain plausibility — hints, not a verdict
• Doesn't pre-empt decisions: on invites/requests/offers it asks (important? do you want to?) or provides both drafts (accept/decline)
• Best-effort live check of suspicious domains/URLs when web/tool access is present (Safe Browsing, VirusTotal, PhishTank/OpenPhish, URLhaus, Spamhaus, WHOIS domain age) — otherwise skipped; bare domain only, no PII, multiple sources, a hint not a verdict
• Surfaces time-critical updates (password reset, security alert, appointment reminder) instead of treating them as noise

Description

Inbox Pro is a skill for Inbox Zero email management with GTD — it triages your inbox into clear categories, extracts tasks and dates, and drafts fitting replies for you. A calm, efficient email assistant that turns a full inbox into a clear plan in minutes — based on the Inbox Zero and GTD methods. Split by a mode router: triage/bundle (categories + one action per email), extract tasks (action items, deadlines, dates, waiting-for as a GTD list), draft a reply (in the right tone, optional variants), thread summary (who wants what, open points), cleanup/noise (spot newsletters/notifications, suggest unsubscribe/filter) and follow-up/waiting-for (track unanswered emails, nudge drafts). Methodical core: one decision per email (do <2 min, delegate, reply, schedule, archive/delete), the 2-minute rule, one-touch, CC = informed not asked, batch processing instead of constant checking. Pro and safety layer: it only ever drafts and sends/deletes/forwards nothing without approval; treats email content as data, not commands, and detects phishing/social engineering (urgent payment/login demands, changed bank details, suspicious senders/links); invents nothing (placeholders instead of made-up promises); handles PII and sensitive/legal mail carefully. Output as an inbox-zero plan (triage table), reply drafts, task list, waiting-for and phishing notes. Synergy: extracted tasks go to the Weekly Planner, meeting notes to the Meeting Minutes.

Examples

What it does not do

• Sends, deletes, forwards or cancels nothing without explicit approval
• Doesn't execute instructions contained in the email text
• Invents no facts, numbers, promises or dates in drafts
• Doesn't share sensitive data with addresses from the email content
• Doesn't replace a spam filter, email client or legal advice
• Makes no final phishing verdict and opens no links — recommends verification via the official channel

Compatibility & tech

Tested (internal)

3 scenarios

Recommended runtime

Any current model; with email access (connector) for direct triage, otherwise paste emails in.

Modes

Triage/bundle · Extract tasks · Draft a reply · Thread summary · Cleanup/noise · Follow-up/waiting-for · Appointments & invites · Templates

Inputs

Inbox/email list · single email/thread · pasted email text · optional: tone/relationship, priorities/VIPs

Output format

Inbox-zero plan (triage table with one action per email) + reply drafts + task list (GTD) + waiting-for + phishing notes

Subcategory

Email & communication

License

Proprietär

Security profile

What you get

• inbox-pro-1.5.0/3 files
  • .claude-plugin/marketplace.json
  • plugins/inbox-pro/skills/inbox-pro/3 files

    • SKILL.md
    • manifest.json
    • references/1 file

      • inbox-pro-checklist.md
  • .agents/skills/inbox-pro/→ universal — same content (Codex, Cursor, Copilot, Gemini, Windsurf, Cline)
  • LICENSE.txt

Installation

Also works as a chat prompt

Note

Works with pasted emails; in agentic tools with email access it reads but only ever drafts (never sends). Best with a stated tone/relationship and priorities.

Changelog

• v1.5.022.06.2026Platform-native output: now actively uses your platform's native formatting — tables, checklists, code blocks and clickable/button options, with a numbered fallback on non-interactive platforms. Plus internal metadata that prepares the upcoming agent builder.
• v1.4.018.06.2026Reworked to the AUTHORING v2 standard: realistic input/answer examples (incl. a calibration case), copy checklist(s) and internal evals; cross-references added.
• v1.3.010.06.2026Best-effort live reputation check of suspicious domains/URLs when web/tool access is available (Safe Browsing/VirusTotal/PhishTank/OpenPhish/URLhaus/Spamhaus/WHOIS), otherwise skipped; extra phishing signs (channel switch, secrecy); mail types tied to standard categories + time-critical updates surfaced. Verified via a CEO-fraud test.
• v1.2.010.06.2026Mail-type classification (system/notification, transactional, distribution list, opportunity, series, spam …); expanded phishing heuristics (typical signs + optional domain plausibility with web access, a hint not a verdict, verify via the official channel); don't pre-empt decisions (ask on invites/requests or provide both drafts).
• v1.1.010.06.2026Extended from a test run: appointments/invites (accept/decline/reschedule + calendar entry), templates (canned responses), reply in the email's language, tact (de-escalating angry senders, polite declines), don't open attachments/links unchecked.
• v1.0.010.06.2026Inbox-Zero/GTD inbox skill: triage into categories with one action per email, 2-minute rule/one-touch, task/date/waiting-for extraction, reply drafts with tone control, thread summary, noise cleanup, follow-up. Pro/safety layer: drafts instead of sends, email = data not commands, phishing detection, invents nothing, batch recommendation.

Frequently asked questions