Correct Vale writing errors with AI
Resolve error-level Vale alerts without breaking code examples or changing technical meaning. Review each document in its own pull request.
Correct Vale writing errors with AI when a documentation change mixes prose and working examples. The fix skill identifies the rule behind an alert and checks whether its recommended wording fits the sentence. A command option inside a code block, for example, must keep the spelling the program accepts.
The default scope is error-level alerts in the files you selected. The agent checks each changed file again, keeps the markup intact and prepares a separate pull request. If an edit would change what a reader is being told to do, the reviewer gets the line and the reason it was left for a person.
What it does
- Use the rule’s actual replacement when it preserves the intended instruction.
- Keep code fences, reference links, MDX expressions and existing line breaks intact.
- Record the rule and the location for each accepted correction.
- Leave an ambiguous sentence for its author and explain the issue.
- Re-run Vale on each edited document before preparing its review.
Install and use
- Install the skill with Vale, jq and Git available. Opening reviews needs permission to branch and submit pull requests to the chosen repository.
- Ask “Correct the errors in these two guides. Keep the examples intact and explain any change that needs an author.”
- Read the recheck and compare each file in its own review.
FAQ
Source
Made by Joseph Kato (Vale). Original skill: fix, published on GitHub under the MIT licence.
View on GitHub