Classify

Step 2: Read the Task

Call get_issue with the issue ID to read the title and description.

Step 3: Decide

Simple — ALL of these are true:

• Clear, specific scope (e.g. "fix typo on settings page", "add loading spinner")
• No architectural decisions needed
• Likely touches 1-5 files
• No ambiguity in what needs to be done

Complex — ANY of these are true:

• Vague or broad scope (e.g. "build user management", "migrate to new API")
• Requires architectural decisions
• Likely touches 10+ files or multiple systems
• Needs user discussion to clarify requirements
• Could be broken into multiple independent tasks

If unsure, choose complex.

Step 4: Execute

If Simple

Execute these tool calls in order:

1. add_issue_tag — add relevant domain tags from the tags returned by get_context in Step 1. Only use tags that actually exist in the project — never invent tag names.
2. update_issue — append autopilot metadata to description and set priority:
   • Append to existing description:

     <!-- autopilot
     tier: {low|medium|high}
     agent: {agent type}
     -->
     

   • Set priority field
   • Set status_id to the "To Do" status ID
3. Output one line: Classified as simple → To Do | tags: {tags} | tier: {tier} | agent: {agent}

If Complex

Execute these tool calls in order:

1. add_issue_tag — add the brainstorm tag (find its ID from get_context)
2. update_issue — set status_id to the "Triage" status ID
3. Output one line: Classified as complex → Triage

That's it. Done. No follow-up, no summary, no next steps.

Model-Agent Reference (for simple tasks only)