What it does

Scans LLM-generated output (code, suggestions, docs) for references to files, functions, classes, or packages that do NOT actually exist in the project.

Use BEFORE pasting AI suggestions into your editor, reviewing AI-generated PRs, or trusting code that claims "as shown in foo.

How it arrives

You don't install skills one by one. ClevskillSetup installs the whole curated pack, wires the safety hooks, and verifies each skill actually runs on your machine. The weekly health scan re-checks it from then on, and repairs route through your approval.

Setup

No setup needed. It works as soon as the pack is installed; no accounts, no API keys, nothing leaves your machine.

Invoke it

In Claude Code or Claude Cowork, just ask in plain language; skills trigger on intent. Things people say that make this skill run:

you: did the AI hallucinate
you: verify these file paths
you: check for fake imports
claude: [hallucination-detector runs, result returned with full audit log]

Part of the security suite: it runs automatically where it applies, fails closed on errors, and never sends your content anywhere.