What is the correct way to run shell scripts in a hook in v9?

[janeklb]

In v8 etc we could put anything in the hook file

eg.

#!/bin/sh
. "$(dirname "$0")/_/husky.sh"

echo "Doing step 1 of pre-commit hook"
npx ....


echo "Doing step 2 of pre-commit hook"
npx ...

What's the correct way to achieve this in v9?

If i remove the shebang header and . "$(dirname "$0") ..., then the hooks fails because it doesn't understand echo "...". If I put the shebang back in then seems to work ok..

#!/bin/sh

echo "Doing step 1 of pre-commit hook"
npx ....


echo "Doing step 2 of pre-commit hook"
npx ...

However, if I remove the _ directory (rm -rf .husky/_) and run npm run prepare (initialising husky), then the _ is re-created. Should I be sourcing this directory in the hook script (as before), or is it purely optional now?

[ignaciobockl]

Hi @janeklb

In Husky v9, hooks are plain POSIX shell scripts — no shebang or sourcing needed.

What changed in v9

The .husky/_ directory is a Husky internal directory. You don't need to source it. Hook scripts are simple shell commands:

echo "Step 1"
npx lint-staged

echo "Step 2"
npx ...

No shebang required

In v9, Husky executes hooks directly. The shebang and the sourcing line from v8 are no longer needed.

If you need Bash features

If your hook requires Bash-specific syntax, use this pattern:

bash << EOF
# Your bash script here
if [ "$SOME_VAR" = "value" ]; then
  echo "Conditional step"
fi
EOF

Testing

After editing, test with:

git commit --allow-empty -m "test hook"

The exit 1 trick still works to test without committing.