Agent Sorting Hat

Use this skill when an existing agent or agent team feels wrong, overloaded, blurred, under-specified, or misaligned.

This skill is for sorting / alignment / placement. It is not for final persona polishing, identity-binding, or character-soul refinement.

If the problem is “this agent seems off,” “this team boundary is unclear,” “should this be one agent or three,” or “this role keeps drifting,” use this skill before rewriting persona files.

Use this skill when

Trigger on requests like:

• “This agent feels wrong.”
• “She keeps giving plans but not executing.”
• “This persona and its job seem mismatched.”
• “We need to rebuild this agent.”
• “These agents have blurry boundaries.”
• “Should this be split into multiple roles?”
• “Do we need a new agent, or is the current one just in the wrong position?”
• “Our team structure is off.”

Do not use this skill for

Do not use this skill as the first choice when:

1. The user only wants persona refinement, tone reshaping, identity-binding, or stronger character feel.
   • Use ollivanders-agent-shop instead.
2. The user wants a brand-new agent from scratch with no role-drift or team-structure question.
   • Use the most appropriate creation/persona skill instead.
3. The user only wants a small wording tweak to one file.
   • Edit directly if no deeper role problem exists.

Core rule

Diagnose placement before personality.

Do not jump into persona rewriting until you have checked whether the real problem is:

1. Role misalignment — the agent is in the wrong job.
2. Team gap — one agent is covering too many incompatible responsibilities.
3. Operating problem — the role is correct, but the workflow, subagent use, delegation judgment, or output pattern is broken.
4. Persona problem — the role is correct, but the tone, identity, or character grounding is weak.

Use this order strictly:

1. Role misalignment
2. Team gap
3. Operating problem
4. Persona problem

Required workflow

Follow these steps in order.

1. Read the current role context

Read only the files needed to understand the current situation, typically:

• the target agent’s role/persona files
• any relevant team/proposal docs
• any current-team-state or registry documents
• one or two strong comparison examples, if relevant

2. Identify the task chain

Determine the actual work chain behind the problem.

Typical chain examples:

• research / analysis
• design / strategy
• execution / production
• audit / risk
• quality control
• coordination / integration

Do not start by asking “who is this character?” Start by asking: what work chain is actually happening here?

3. Diagnose the problem type

Classify the issue into one or more of:

• role misalignment
• team gap
• operating problem
• persona problem

State clearly which one is primary.

4. Decide whether to keep, split, merge, or reserve

For each affected role, determine whether it should: