Which repo-defined commands do you check before opening a project?

I'm building ClonePeek, an open-source Python CLI and GitHub Action that inventories the execution hooks a repository can define: npm lifecycle scripts, VS Code folder-open tasks, devcontainer commands, Cursor/Claude Code hooks, and local MCP server launch commands.

A regression test surprised me: our "read-only" Git changed-files check could invoke a locally configured fsmonitor helper. We now constrain the Git subprocess as well as passively reading project config. That bug changed how I think about scanning untrusted checkouts.

ClonePeek is intentionally a heuristic review aid, not malware detection or a sandbox. It's easy for a clean scan to miss dynamically constructed commands; I don't want a false sense of security.

What is one harmless but real repo hook you'd want a reviewer to notice? Or one that scanners like this would probably misclassify? I'm especially interested in reproducible, inert examples and feedback before the Oct 20 launch.

Maintainer disclosure: I built ClonePeek. Source, tests, and limitations are public on the product's GitHub repository.