Problem

The test-proxy.md workflow was creating a new GitHub issue every time it ran, which clutters the repository with test result issues. This behavior was problematic because:

• Issues were created on every workflow run (both successful and failed tests)
• The issues didn't provide actionable feedback in the context where they were most useful
• Repository maintainers had to manually close these automated test result issues

Solution

Updated the workflow to use the issue comment output processing instead of issue creation:

Before:

permissions:
  issues: write # needed to write the output report to an issue

tools:
  github:
    allowed:
      - "create_issue"
      - "create_comment"
      - "get_issue"

After:

output:
  issue_comment: {}

tools:
  github:
    allowed:
      - "create_comment"
      - "get_issue"

Key Changes

1. Replaced issue creation with comment creation: Added output.issue_comment: {} to enable automatic comment creation on PRs instead of standalone issues

2. Removed explicit permissions: The compiler now automatically handles permissions for the comment creation job, providing appropriate issues: write and pull-requests: write only to the comment job

3. Updated reporting instructions: Modified the natural language prompt to focus on commenting when there are failures or notable issues, rather than always creating comprehensive reports

4. Conditional execution: Comments are only created when the agent produces output and the workflow runs in a PR context

Benefits

• Contextual feedback: Test results now appear as comments on the relevant PR where the changes are being reviewed
• Reduced noise: No more standalone issues cluttering the repository
• Conditional reporting: Comments are only created when there's meaningful output to share
• Proper permissions: The main workflow job no longer needs write permissions, improving security

The compiled workflow now generates a create_issue_comment job that runs conditionally after the main test job, providing feedback directly where it's most useful.

💡 You can make Copilot smarter by setting up custom instructions, customizing its development environment and configuring Model Context Protocol (MCP) servers. Learn more Copilot coding agent tips in the docs.