Aicheck — AI skill for Claude Code

aicheck

Know what to check before you trust an AI answer. Start a message with `aicheck`:

aicheck for my essay

The assistant reviews the answer (the one you paste, or its own previous reply) and lists the few claims that could be wrong, how risky each one is, and exactly where to check it. It flags risks. It never pretends to confirm them from memory.

AI answers sound equally sure when they're right and when they're wrong. `aicheck` shows you where to look.

What you get

**Checking:** three claims about remote-work productivity and burnout
**Used for:** an essay you will submit
**Stakes:** high: an invented source in a submitted essay costs marks and trust

| # | Claim                                         | Risk | Check it by                                      |
|---|-----------------------------------------------|------|--------------------------------------------------|
| 1 | HBR article "The Hybrid Paradox", 22% figure  | 🔴   | Search the exact title on hbr.org                |
| 2 | Stanford study: remote staff 13% more productive | 🔴 | Open the original 2015 paper, find the 13%       |
| 3 | "Most companies now offer hybrid work"        | 🟡   | Find one named survey with a year, or cut it     |

**Check first:** #1, because if the article doesn't exist, the paragraph has no source.
**Safe to use as-is:** nothing until #1 and #2 are checked.

Next? (check it for me / fix the answer / done)

Reply `check it for me` and, if your AI can browse, it checks the red rows against real sources and links them. Reply `fix the answer` to get the answer rewritten with the risky claims that are still unchecked marked `[unverified]`. See [`examples/`](examples/) for a code check, a real news summary checked before investing, and a poem it refuses to check.

What is in this repo?

File	Use it for