draft-message-desk

Review and harden outbound message drafts against source material before sending.

Updated Sep 11, 2026
One-click install
npx skills add https://github.com/JYeswak/grok_bot_playground --skill draft-message-desk-jyeswak
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: draft-message-desk
Source: https://github.com/JYeswak/grok_bot_playground/tree/main/plugin/skills/draft-message-desk
Command: npx skills add https://github.com/JYeswak/grok_bot_playground --skill draft-message-desk-jyeswak

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Outbound messages often go out with unverified claims, unresolved placeholders, or wording the owner never approved, and a sent message cannot be taken back. This Skill turns a rough intent into one exact, source-checked draft that stops at review and never sends anything on its own. ## Core Features & Use Cases - Exact draft production: Produces one complete wording with subject, body, and sign-off, with no blanks, bracketed fill-ins, or unresolved alternatives. - Claim verification: Traces every factual statement, date, price, or quote back to the stated intent or a named source, cutting anything that does not trace. - Send-lock boundary: Never sends, schedules, or publishes; only the owner's explicit yes on that exact message authorizes a send, and relayed approval does not count. - Use Case: After a sales call, ask for a follow-up email draft. The Skill produces the exact wording, a claim table citing the call transcript for each statement, and a reminder that the draft stays unsent until you approve it. ## Quick Start Ask the assistant to draft a follow-up email to a named recipient based on your meeting notes, and request the claim table showing the source for every statement before you approve anything.

Frequently Asked Questions about draft-message-desk

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I review an email draft before sending it?

Provide the message intent, the recipient and channel, and any source material such as notes or transcripts. The Skill returns one exact draft plus a claim table showing the source for each statement, and the draft stays unsent until you explicitly approve that exact wording.

How to check that claims in an outbound message are accurate?

Each factual statement, date, price, or quote is traced back to the stated intent or a named source such as a thread, transcript timestamp, or page. Claims without a cited source are cut rather than softened, so nothing invented reaches the draft.

Can this skill send or schedule the message it drafts?

No. It drafts only and never sends, schedules, sequences, or publishes. Sending requires the owner's explicit yes on that exact message as a separate step, and approval relayed through another person or thread does not count.

What happens when a source document is unreachable?

A source that cannot be read is recorded as unreachable and that line of the draft stops, rather than being improvised around. If a source needs a credential that was not provided, it is noted and the related claims are not drafted.

When should I not use a draft review workflow?

Do not use it for inbound triage, deciding who to contact, list building, or prospect research, since those belong to their own routines. It also does not handle anything that sends or publishes, which requires a separate owner-approved step.