manuelschipper/nah
A permission guard for coding agents. It sorts each action by what it does, such as deleting files, sending data out or rewriting git history, rather than by command name, then allows, asks or blocks it under fixed rules. An optional model step handles only commands it does not recognise.
Idea adaptedKey files read, not run
I learnt this repo's idea and adapted it into my own workspace setup.
How we know
We read the README and key files, not every source file, and did not run it, so the most we can say is Idea adapted or Risky.
Good for
Developers who let an agent approve its own commands but still want clear rules on what it may do.
Not for
A full sandbox: it guards at the command level. The benchmark is the author's own, on one set of bash calls, and the project is brand new.
Repo
- Stars
- 481
- License
- MITUse it, change it, sell it, closed-source included. Keep the author's copyright notice.
- Maturity
- active
- Last push
- 2026-09-10
- Last checked
- 2026-11-14