Autofix provides you with targeted recommendations to help you fix code scanning alerts so you can avoid introducing new security vulnerabilities. The potential fixes are generated automatically by large language models (LLMs) using data from the codebase and from code scanning analysis.
How autofix works
Autofix translates the description and location of a code scanning alert into code changes that may fix it. It interfaces with the large language model GPT-5.3-Codex from OpenAI, which has sufficient generative capabilities to produce both suggested fixes in code and explanatory text for those fixes.
There are two ways to get a fix for an alert: agentic autofix and Autofixo do Copilot. If agente de nuvem Copilot is available in a repository, assigning an alert uses agentic autofix instead of Autofixo do Copilot.
Agentic autofix
Observação
This feature is currently in public preview and is subject to change.
Assign a code scanning alert to Copilot to have it resolve the alert for you. Assigning an alert starts an agent session: agente de nuvem Copilot calls tools to explore your codebase beyond the affected file, generates a fix, validates it (for example, by re-running CodeQL), and iterates until it opens a pull request with the changes. See Resolver alertas de varredura de código.
Keep the following in mind:
- Agentic autofix requires agente de nuvem Copilot and Autofixo do Copilot to be available in the repository. If agente de nuvem Copilot isn't available, assigning an alert falls back to Autofixo do Copilot instead.
- Each agentic autofix session is billed as a agente de nuvem Copilot session and consumes AI credits. See Sobre o agente de nuvem do GitHub Copilot.
- Copilot follows any custom instructions configured for the repository or organization when it generates a fix.
- Agentic autofix works on a best-effort basis. Copilot validates fixes by re-running CodeQL using the code-scanning query suite, so it can't confirm that a fix resolves alerts generated by custom queries or the security-extended query suite. Fix quality for alerts from third-party tools is also not guaranteed.
Getting a suggested fix with Autofixo do Copilot
Autofixo do Copilot generates a single suggested fix for an alert, which you review and apply yourself.
You do not need a subscription to GitHub Copilot to use Correção Automática do GitHub Copilot, and it does not consume AI credits. Autofixo do Copilot is available to all public repositories on GitHub.com, as well as internal or private repositories owned by organizations and enterprises that have a license for GitHub Code Security.
Autofixo do Copilot is allowed by default and enabled for every repository that uses CodeQL, regardless of whether it uses default or advanced setup for code scanning. There is no separate step to enable Autofixo do Copilot: enabling code scanning with CodeQL is sufficient. See Como definir a configuração padrão da verificação de código.
Administrators at the enterprise, organization, and repository levels can choose to disable Autofixo do Copilot. If Autofixo do Copilot has been disabled at your level, you can re-enable it by following the same steps used to disable it and selecting the option to allow Autofixo do Copilot. To learn how to manage Autofixo do Copilot at each level, see Desabilitar a correção automática do Copilot para alertas de segurança de verificação de código.