name: reference-data-builder description: > Build a Client Reference Data document from onboarding information. Takes client details, team roster, tools, and engagement info as input and produces a complete reference data file ready to upload to all agents. Use when: "build reference data", "set up a new client", "create client reference file", "onboard a new client", or any request to create the canonical client data document. metadata: author: Kathryn Brown, Advisory OS version: "1.0.0" updated: "2026-04-26"
Client Reference Data Builder
Build the canonical source of truth for a new client in under 10 minutes. Every proper noun, every team member, every tool — documented once, used by every agent.
Core Principle
One document, every agent. The reference data file is uploaded to every agent project for this client. If a name appears differently in a transcript than in this document, this document wins. Build it right once. Update it when things change. Never let agents guess at spellings.
What This Skill Does
Paste whatever you have — onboarding notes, intake forms, emails, session transcripts, LinkedIn profiles, or just tell me what you know about the client. You'll get back a complete Client Reference Data document ready to upload.
Job 1: Extract and Organize — Pulls every proper noun, team member, tool name, and engagement detail from your input and organizes them into the standard reference data structure.
Job 2: Flag Spelling Risks — Identifies names and terms likely to be misspelled by audio-to-text transcription and pre-populates the correction mapping.
Job 3: Produce Upload-Ready Output — Generates the complete markdown document in the exact template format. Copy, save, upload to agents.
The Reference Data Document: Section by Section
1. Company
What it contains: Legal name, common name (if different), industry, location/timezone.
Why it matters: Every agent needs to know who the client is and where they operate. Timezone affects scheduling references in all outputs.
Format: Two-column table (Field | Value).
2. Team Roster
What it contains: Every person who appears in sessions, constraint briefs, or will be named in builds. Correct spelling, role, and transcript misspelling notes.
Why it matters: Audio transcription mangles names. "Jennifer" becomes "Jenna." "Steinberg" becomes "Sternberg." This table is the override.
Format: Three-column table (Name | Role | Notes). Notes column flags known transcript misspellings.
3. Tools & Software
What it contains: Every tool the client uses, what it's used for, and how transcripts misspell it.
Why it matters: "Financial Cents" becomes "Financial Sense" in transcripts. Agents that reference the wrong tool name lose credibility in builds.
Format: Three-column table (Tool | What it's used for | Transcript misspelling).
4. Key Proper Nouns
What it contains: Catch-all for anything else that gets misspelled — client names, vendor names, software features, industry terms.
Why it matters: The first time an agent writes "Quickbooks" instead of "QuickBooks" in a client deliverable, trust drops.
Format: Three-column table (Correct | Incorrect | Context).
5. Engagement
What it contains: Entry point, current engagement type, rate, start date, meeting cadence, advisor name, brand.
Why it matters: Agents reference engagement context when producing session recaps, constraint briefs, and project plans. Wrong engagement details produce wrong output.
Format: Two-column table (Field | Value). Advisor defaults to Kathryn Brown. Brand defaults to Advisory OS.
6. Update Log
What it contains: Date, what changed, who updated it.
Why it matters: When a team member leaves or a tool changes, this log tracks when the reference data was updated so agents always know which version they're running.
Format: Three-column table (Date | What Changed | Updated By). First entry is "Initial build" with today's date.
Rules
- One client per run. Don't combine multiple clients in a single input. Each client gets their own reference data document.
- Extract, don't assume. Only include information explicitly stated in the input. If a role isn't mentioned, leave the role column with "[confirm]" — never guess.
- Flag transcript risks proactively. If a name has common misspellings (e.g., "Catherine" vs. "Kathryn," "Schneider" vs. "Snyder"), pre-populate the Notes column even if the input doesn't mention the misspelling.
- Preserve exact spelling from the input. If the input says "Steinberg" — use "Steinberg." Don't normalize to a more common spelling.
- Mark unknowns, don't skip them. If the input doesn't include a field (e.g., timezone, rate), include the row with "[unknown — confirm with client]" rather than omitting it. Every section appears in every output.
- Include the usage rules. Every output includes the USAGE RULES footer. These are operational instructions for agents, not decoration.
Output Format
# CLIENT REFERENCE DATA
## [Client Name] — [Company Name]
*Created: [Today's Date] | Last Updated: [Today's Date]*
*This document is the canonical source of truth for all proper nouns, team members, and tools referenced across Advisory OS agents. Upload this file to every agent project when processing this client. All agents override transcript/JSON spellings with this document.*
---
## Company
| Field | Value |
|-------|-------|
| Legal Name | [value] |
| Common Name (if different) | [value or "—"] |
| Industry | [value] |
| Location / Time Zone | [value] |
---
## Team Roster
| Name (correct spelling) | Role | Notes |
|--------------------------|------|-------|
| [name] | [role] | [transcript misspelling notes or "—"] |
*Include: anyone who appears in sessions, constraint briefs, or will be named in builds. Flag transcript misspellings in Notes column so agents know what to override.*
---
## Tools & Software
| Tool (correct name) | What it's used for | Transcript misspelling |
|----------------------|--------------------|----------------------|
| [tool] | [usage] | [misspelling or "—"] |
---
## Key Proper Nouns
| Correct | Incorrect (from transcripts) | Context |
|---------|------------------------------|---------|
| [correct] | [incorrect or "—"] | [context] |
*Catch-all for anything else that gets misspelled in audio-to-text: client names, vendor names, software features, industry terms.*
---
## Engagement
| Field | Value |
|-------|-------|
| Entry Point | [value] |
| Current Engagement | [value] |
| Rate | [value] |
| Start Date | [value] |
| Meeting Cadence | [value] |
| Advisor | Kathryn Brown |
| Brand | Advisory OS |
---
## Update Log
| Date | What Changed | Updated By |
|------|-------------|------------|
| [Today's Date] | Initial build | [Advisor name] |
---
### USAGE RULES (for all agents)
1. **This document overrides all other sources** for proper nouns — transcripts, JSONs, constraint briefs, matrix outputs. If a name appears differently in a transcript than in this document, this document wins.
2. **Upload to every agent** when processing this client: IC System, Build Design Agent, AISOP Factory, QC Agent, Constraint Priority Matrix.
3. **Created during onboarding.** Advisor fills this out before the first agent run for a new client.
4. **Updated when team changes.** New hire, departure, role change, new tool adopted — update this document first, then re-upload to active agent projects.
5. **Layer 1 Reference section mirrors this document.** When the IC system builds Layer 1, it pulls the Reference block from this file. They should always match. If they don't, this file is the authority.What Makes This Different
You could fill out the reference data template by hand. Many advisors do — and they miss things.
The Reference Data Builder extracts information from whatever you have. Onboarding notes, intake forms, email threads, transcripts, LinkedIn profiles — paste it all. The skill organizes it into the standard structure, flags spelling risks you wouldn't catch, and produces a document that every agent in the Advisory OS stack can consume.
The difference isn't convenience. It's completeness. The skill catches the proper nouns you'd forget to include until an agent misspells them in a client deliverable.
License
Copyright (c) 2026 Kathryn Brown, Practice Builders
This skill is licensed for your personal and business use. You may run this skill inside your own practice and share the outputs it produces with your team and clients. "Your practice" includes employees and contractors engaged to perform work for your business under your direction — virtual assistants, operations support, bookkeepers, and similar team members.
You may not share, distribute, resell, or repackage the skill file itself — including this SKILL.md document, its prompts, frameworks, and structure — with anyone outside your practice. This includes peer practitioners, other consultants who would use it in their own client work, and anyone outside your operating team. Written permission from Kathryn Brown (kathryn@creatingyourplan.com) is required for any redistribution.
This skill is provided "as is" without warranty of any kind, express or implied.