Prevent AI Agents from auto-merging PRs
Came across this post from Elvis containing a helpful tip to prevent your AI agent from accidentally auto-merging PRs when using the --YOLO or --dangerously-skip-permissions mode. It blocks AI agents like Claude, Codex, or OpenCode from merging Pull Requests accidentally.
In simple language, this setup checks when your AI agent runs a command containing pr merge and:
- if it does match: blocks the action and prints and error message
- if it does not match: allows running it without errors
This reduces the risks when vibe-coding in YOLO mode, and here's how I have set this up:
# make the hidden folder
mkdir -p ~/.local/bin
# create and open new file
nano ~/.local/bin/gh
And then paste the following in the file:
#!/bin/bash
# Wrapper to block AI agents from merging PRs
if [[ "$*" == *"pr merge"* ]]; then
echo "🚫 Blocked: merging requires human approval"
exit 1
fi
exec /opt/homebrew/bin/gh "$@"
Now, make the script runnable:
chmod +x ~/.local/bin/gh
Then I added the following in my .zshrc file:
echo 'export PATH="$HOME/.local/bin:$PATH"' >> ~/.zshrc
Reloaded settings using the following command:
source ~/.zshrc
And it was ready.
It will now block commands containing pr merge in them, in case my AI agent starts to go rogue.
Webmentions