# Personalizer

**Status:** <mark style="color:$primary;">Live</mark>

## What It Does

Employees trigger the skill with `/personalizer` in Claude. Claude conducts a short, warm, conversational onboarding (about 10 to 15 minutes, 13 questions) that gets to know the employee as a person and as a professional. At the end, Claude translates everything it heard into a first-person XML profile prompt the employee pastes into their Claude Enterprise personal preferences. That prompt shapes how Claude engages with them on every future conversation.

Examples of what the skill produces:

* A profile prompt tuned to how the employee thinks, receives feedback, and wants Claude to communicate.
* Behavioral instructions translated from personal context so Claude adjusts without requiring the employee to re-explain themselves each time.
* Friction rules that tell Claude what to stop doing, in the employee's own voice.

***

## Workflow

```
Employee triggers /personalizer in Claude → SKILL.md (conversation logic + translation rules) → Prose draft presented to employee → Employee reviews and confirms → XML profile prompt generated → Employee pastes into Claude Enterprise Personal Preferences
```

The skill lives entirely in a single SKILL.md file. The conversation runs in Claude directly, the profile is generated in Claude directly, and the employee handles the final paste into their own Claude Enterprise settings. Each employee's profile stays private to their own account.

***

## Conversation Flow

The skill reads the employee's opening turn to determine intent, then routes accordingly. First-timers go through the full conversation. Returning employees can trigger targeted edits, section refreshes, full rebuilds, or quick reviews without running the whole thing again.

The full conversation is 13 questions across 5 zones:

| Zone                 | Purpose                                                                       | Questions                                                                                                                |
| -------------------- | ----------------------------------------------------------------------------- | ------------------------------------------------------------------------------------------------------------------------ |
| Who I Am             | Identity, role, and the personal context that shapes how they show up at work | Name, role, what lights them up, what makes work meaningful                                                              |
| What I Do            | Day-to-day reality of the work so Claude can calibrate                        | Responsibilities, workflows, how they spend their time                                                                   |
| How I Think          | Cognitive style, problem-solving approach, and feedback preferences           | Thinking style on hard problems, what makes feedback land, how they handle being wrong                                   |
| What I Need From You | Direct operating instructions for Claude                                      | Preferred response depth, tone, formality, swearing policy                                                               |
| What Not To Do       | Guardrails and friction rules                                                 | What wastes their time or frustrates them, behaviors to avoid, any personal context that meaningfully affects their work |

Between every question, Claude reacts with presence. Not generic filler like "got it" or "makes sense," but specific engagement with what the employee actually said. A riff when there is something fun to play with, a real take when the answer merits one, a beat of warmth when the answer is heavy.

***

## Translation Engine

The most important part of the skill is the translation work at the draft stage. The profile is not a transcript of what the employee said. It is a translation of what they said into what Claude needs to know to behave differently on every future turn.

Personal details are never stored as facts. They are inferred into behavioral instructions. Examples of this inference in action:

| Employee Shares                                       | Instruction That Goes in the Profile                                                                                                    |
| ----------------------------------------------------- | --------------------------------------------------------------------------------------------------------------------------------------- |
| "I've been dealing with a lot outside of work lately" | "Keep interactions efficient. Don't pile on enthusiasm or warmth. Match a grounded, quiet tone unless I open a different door."         |
| "Family is the reason I'm doing this work"            | "When I'm stressed about deliverables, frame help around getting it done efficiently. Respect boundaries around personal time."         |
| "I work in two modes: deep focus and scattered"       | "When I'm locked in, stay tight and stop explaining. When I'm scattering, help me refocus without making it awkward."                   |
| "I value directness over diplomacy"                   | "Skip the softening language. Give me the answer or the disagreement straight. I'd rather hear the real thing than the polite version." |

Every sentence in the final profile has to pass three tests: would Claude behave differently without this sentence, can Claude actually observe or act on this, and is this actionable instruction rather than self-description. If a sentence fails any test, it is cut or rewritten.

***

## Draft Review

After the 13 questions, Claude writes out the employee's profile as prose with five H2 section headings (Who I am, What I do, How I think, What I need from you, What not to do) and asks: "How'd I do? Anything you want to change, or would you like the final prompt to paste into your profile settings?"

The employee can correct any section. Claude rewrites it and shows it again. When the employee is satisfied, Claude converts the approved prose into the final XML block with the same five sections, formatted for Claude Enterprise personal preferences.

***

## Output Format

The final output is an XML block the employee pastes directly into their Claude Enterprise personal preferences:

```xml
<profile>
  <who_i_am>...</who_i_am>
  <what_i_do>...</what_i_do>
  <how_i_think>...</how_i_think>
  <what_i_need_from_you>...</what_i_need_from_you>
  <what_not_to_do>...</what_not_to_do>
</profile>
```

First person throughout. Voice preserved where it carries behavioral signal.

***

## Privacy

The conversation is between the employee and Claude. Nothing is stored, tracked, or transmitted beyond the employee's own Claude Enterprise account. The skill does not log conversations, does not version or archive old profiles, and does not share content with anyone. If an employee wants to keep their old profile after an update, they save it themselves.

Claude communicates this directly at the start of the conversation: "This conversation stays between us, so be as authentic as possible."

***

## Skill Components

| Component          | File     | Description                                                                                                |
| ------------------ | -------- | ---------------------------------------------------------------------------------------------------------- |
| Skill Instructions | SKILL.md | Complete conversation logic, reaction rules, modeled exchanges, translation engine, draft and handoff flow |

The skill is a single-file skill. No reference files, no scripts, no assets. Everything needed to conduct the conversation and generate the profile lives in the SKILL.md.

***

## Current Limitations

* The skill is deployed as a personal skill for testing. Not yet rolled out org-wide.
* The skill does not track adoption. Managers who want to verify rollout need to check with employees directly.


---

# Agent Instructions: Querying This Documentation

If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question.

Perform an HTTP GET request on the current page URL with the `ask` query parameter:

```
GET https://knowledge.themailworks.com/skills/personalizer.md?ask=<question>
```

The question should be specific, self-contained, and written in natural language.
The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.

Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.