1. Check your API Token Permissions (Most Common)

If you are running this from a GitHub Actions workflow using the default GITHUB_TOKEN, it likely lacks the required organizational permissions to invoke the Copilot agent bot.

Switch your authentication to a Personal Access Token (PAT) or a GitHub App token with write permissions for both Issues and Copilot Agents, and try the request again.

2. Verify the Assignee Username

GitHub recently standardized the agent naming conventions. If you are forcing "copilot-swe-agent[bot]" inside the assignees array, the API might validate the issue creation but fail to trigger the agent backend.

Try removing the assignees array entirely and let the agent_assignment block handle the routing implicitly:

    { 
  "title": "issue title", 
  "body": "issue body", 
  "labels": [ "agent" ], 
  "agent_assignment": { 
    "target_repo": "my-organization/my-repo", 
    "base_branch": "main", 
    "custom_instructions": "update meta title in header to abc", 
    "custom_agent": "" 
  } 
}

3. Check for Repository Ruleset Blocks

If your organization recently updated branch protections or repository rulesets, the Copilot cloud agent might be getting blocked from starting a session because it can't automatically create its working branch.

Ensure your repository rulesets allow bypass permissions for the Copilot Agent apps.

Hi,

If this was working a few weeks ago and the same payload is now creating the issue but not triggering the Copilot agent, I would first verify whether the problem is with agent assignment itself or with changes to the Copilot agent feature.

A few things to check:

1. Confirm that the repository still has Copilot coding agent enabled and that the required licenses/features are active for the organization.

2. Verify that assigning copilot-swe-agent[bot] through the REST API is still a supported workflow. GitHub occasionally changes preview features or rollout behavior, and an issue can be successfully assigned without actually triggering agent execution.

3. Try creating an issue manually through the GitHub UI and assigning the Copilot agent. If the agent starts correctly from the UI but not through the API, that would strongly suggest an API-related issue rather than a repository configuration problem.

4. Check the issue timeline and repository audit logs (if available) for any events indicating that the agent assignment was ignored, rejected, or failed to start.

5. Review whether any required fields in agent_assignment have changed since your implementation was first working. Even a newly required field could result in the issue being created successfully while preventing agent execution.

Since your payload already appears to create the issue and assign the bot correctly, it sounds less like a JSON formatting problem and more like either:

• A change in Copilot agent behavior,
• A feature flag/licensing issue,
• Or a regression on GitHub's side.

Could you also share:

• The exact REST endpoint you're calling,
• The API version header,
• And whether manual assignment from the GitHub UI still starts the agent?

That information would help narrow down where the failure is occurring.

Hi,

This kind of issue usually happens when the Copilot Cloud/SWE agent behavior changes on the backend or when the required trigger conditions are no longer fully met, even if the issue itself is created correctly.

A few things to check:

1 - First, verify that the Copilot SWE agent is still enabled for your organization/repository. These features are often opt-in and can be disabled at org level even if API calls still succeed.

2 - The assignee name is important here. In some recent updates, the correct trigger is not always copilot-swe-agent[bot]. In many cases it needs to be the current official Copilot agent handle (for example copilot or the updated GitHub-managed bot identity). If the bot identity changed recently, older values will still assign but won’t trigger execution.

3 - Make sure the repository has Copilot coding agent permissions enabled under settings. Even if assignment works, the agent will not start without proper repository-level access.

4 - Check whether the agent_assignment field is still supported in the same format. GitHub has been iterating quickly on Copilot agent APIs, and fields like custom_agent or target_repo may have changed or been deprecated depending on rollout stage.

5 - Try a minimal payload test:

• Create issue only
• Then assign Copilot manually via API or UI
  If it triggers from UI but not API, it confirms a payload/schema change.

6 - Also check GitHub status and recent Copilot changelogs, because behavior changes (especially for agent execution triggers) are often rolled out silently during preview phases.

Most likely this is not a simple “missing field” but either a permission change or a backend update in how the Copilot agent is triggered from issues. If everything was working a few weeks ago and stopped suddenly, that strongly points to a platform-side change rather than your request body.

If you want, I can help you rewrite the payload in the latest expected format based on current API behavior.