Instantly identify and fix problems in shell scripts with the power of ShellCheck and LLMs.

Access screencast directly here.

See example directory to see script contents before and after.

• ShellCheck should be installed and in your PATH.
• An active OpenAI API key.

brew install kkyr/tap/shellcheck-gpt

Download the latest release and add the executable to your PATH.

go install github.com/kkyr/shellcheck-gpt

Add your OpenAI API key to the environment:

export OPENAI_API_KEY=replace-with-your-api-key

Run shellcheck-gpt against a script:

shellcheck-gpt script.sh

This will:

1. Run shellcheck against script.sh
2. Feed the contents of the script and the output of shellcheck into an OpenAI LLM and ask it to make corrections
3. Write the LLM's output to stdout

If you'd like to instead write the output back into the script, use the -w flag:

shellcheck-gpt -w script.sh

Note that this will ovewrite any contents already in the script.

By default, shellcheck-gpt uses the gpt-3.5-turbo model. You change model using the -m flag:

shellcheck-gpt -m gpt-4-turbo script.sh

See available options and models using the --help flag:

shellcheck-gpt --help

Contributions are welcome!

Areas for improvement:

• Use shellcheck -f diff to apply fixes that can be automatically applied before calling into LLM.
• Support more LLMs
• Add verbose flag to print LLM input/output