Documentation: • Added Copilot agent guidance and usage instructions for the Canary project, including preferred workflows, scoped change practices, validation tips, and a README describing manual and automated assignment flows. Chores • Added an automated workflow enabling authorized maintainers to assign Copilot to fix issues via manual dispatch or a specific issue comment; includes permission checks, input validation, and assignment feedback/comments. |
||
|---|---|---|
| .. | ||
| README.md | ||
Copilot Issue Agent
This repository is configured for GitHub Copilot cloud agent. It does not require an OpenAI API key.
Manual Usage
Repository administrators with GitHub Copilot access can assign an issue to Copilot from the issue sidebar:
- Open the issue.
- Choose
Assignees. - Select
Copilot. - Use
mainas the base branch unless a maintainer wants another branch. - Select the
canary-bugfixagent when available. - Add focused instructions for the bug.
For issue #3552, use:
Fix the highscore pagination bug introduced after PR #3222. Keep the change focused and add or update tests if practical.
Slash Command
After .github/workflows/copilot-issue-fix.yml is on the default branch, repository administrators can comment this exact command at the start of an issue comment:
/copilot fix
Optional extra instructions can be added below the command in the same comment.
The workflow checks that both the actor and triggering actor have repository admin permission before attempting to assign Copilot. It uses GitHub credentials only and does not use OpenAI API secrets.
If GitHub does not allow the default Actions token to assign Copilot in this repository, keep using the manual assignment flow above. A maintainer can later add a narrowly scoped GitHub token as COPILOT_ASSIGN_TOKEN, but that is optional and should only be done if the slash command is worth the additional credential management.