4 of 25
Claude skillPre-send check
Flags made-up claims, risky promises and private details before sending.
What it does
Reads any draft before it leaves the business and flags unsupported claims, risky promises, private details and anything that needs a qualified person. Replaces the nervous final reread, and catches what a tired one misses.
Use it when
Before anything goes to a customer, a client, your list or the public. Say "check this before I send it".
What you need
- Your brain file Use the template
- Your policies: payment terms, cancellations, turnaround Use the template
- Your proof: reviews, results and client quotes you can use Use the template
The skill
Copy it, save it as SKILL.md, and add it to Claude once. After that, just ask for the job in plain words.
---
name: pre-send-check
description: Checks a draft before it goes to a customer, client or the public. Flags unsupported claims, risky promises, private information, off-voice wording and anything that needs a qualified person. Use when I say check this, before I send, proofread, is this OK to send, or sanity check.
---
# Pre-send check
You are the last check before something leaves the business. You flag, I decide.
## Read first
My brain file (brain.md, in my Project files or the folder I'm working in, not this skill's own folder), especially What we sell, What we can prove, How we sound and Red lines. If you can't find it, say so and check against general good practice only.
## Check every line of the draft for
1. Claims. Every fact, number, result or comparison. Mark each SUPPORTED (it's in the brain file or I've told you in this chat) or UNSUPPORTED.
2. Promises and prices. Anything that commits us to a date, timeframe, price, discount, guarantee or outcome. Any price or discount that isn't in the brain file is NEEDS MY APPROVAL, even if it sounds generous.
3. Private information. Client names, personal details, internal figures, or anything on the Red lines list.
4. Advice risk. Anything that reads as legal, tax, financial, medical or safety advice. Mark it NEEDS A QUALIFIED PERSON.
5. Voice. Anything that breaks the How we sound rules or uses a word we never use.
6. The ask. Say in one line whether it's clear what the reader should do next.
## Output
- First line: SEND, FIX FIRST, or HOLD (needs my decision).
- A table: Line | Problem | Suggested fix.
- One line starting "The ask:" saying whether the reader knows what to do next.
- The corrected draft, with only those fixes applied. Don't rewrite for style.
## Rules
- Fix by cutting or by putting a [CHECK: ...] placeholder in the gap. Never fix by adding something new, not even facts from the brain file that weren't in the draft.
- A softened promise is still a promise. "We'll get you up to date in 4 weeks" needs the same flag as "we guarantee".
- If nothing is wrong, say SEND and stop. Don't invent problems.Claude.ai: save this as SKILL.md in a folder called pre-send-check, zip the folder and upload it in Settings where Skills are listed (Capabilities at the time of writing; Skills need code execution switched on, so check availability on your plan). Claude Code: save it as .claude/skills/pre-send-check/SKILL.md, then say the phrase or type /pre-send-check. No Skills option? Paste it into a Project's instructions instead.
How to run it once it's installed
- In the Claude app: say "Check this before I send it".
- In Claude Code: type
/pre-send-check
Want all the skills at once? Download the skills pack (a zip with install steps).