How to automate issue metadata with GitHub issue intents
GitHub walks through using GitHub issue intents to reduce the manual work of keeping issue metadata accurate, while still letting maintainers control what gets applied automatically.
Overview
The problem: issue metadata triage at scale
Maintainers often need to keep incoming issues consistently tagged and prioritized (labels, priority, and other metadata). Doing this manually can become repetitive and time-consuming.
How GitHub issue intents work
GitHub issue intents provide suggestions for issue metadata updates and include:
- Suggested changes (for example, labels and priority)
- A rationale explaining why the suggestion was made
- A confidence signal that can be used to decide how much automation to allow
Controlling automation with confidence thresholds
The video covers configuring repository-level settings to tune how issue intent suggestions are handled, including:
- Setting confidence thresholds
- Choosing when suggestions require review vs when they can be applied automatically
Testing automated labeling and priority suggestions
GitHub demonstrates testing the workflow end-to-end:
- Generating intent-based suggestions for incoming issues
- Reviewing suggested labels/priority
- Validating that the automation behaves as expected before increasing automation levels
Setting up an agentic workflow for issue intents
The video shows how to create an agentic workflow around issue intents so that routine changes can be applied automatically while still supporting human oversight.
Inspecting outputs and logs in GitHub Actions
GitHub covers operational visibility for the automation, including:
- Inspecting safe outputs
- Reviewing agent logs within GitHub Actions to understand what happened and why
Customizing issue intents with Copilot and APIs
The video closes by showing customization options, including:
- Using Copilot to help tailor the workflow
- Using APIs to extend or integrate issue intent behavior with other tooling