05 — OUTPUT SKILL: New Client Kit

The production skill. Contains step-by-step instructions and embedded templates for all three modes.

Mode 1: Day 0 Scaffold

Prerequisites

1. GitHub repo created. Run this command before starting Mode 1:

   gh repo create kathryn-art/aos-client-{{SLUG}} --private --clone

1. All Mode 1 inputs collected (see 01-context.md for the full list).

Step 1: Create Directory Structure

Create the following directories inside the repo root:

aos-client-{{SLUG}}/
├── .claude/
│   └── skills/
│       ├── start/
│       ├── recap-email/
│       ├── recap-internal/
│       ├── recap-json/
│       ├── update-cpm/
│       ├── update-masterplan/
│       ├── update-projectplan/
│       ├── update-blueprint/
│       ├── draft-actions/
│       └── session-agenda/
├── sessions/
├── inputs/
├── drafts/
├── ready/
├── archive/
├── constraints/
├── master-plan/
├── projects/
│   └── non-project-specific/
├── reference/
│   └── core/
├── reference-data/
├── kits/
├── frameworks/
├── brand/
├── skills/
└── qc-agents/

Kanban folders: sessions/ is permanent record (never moves). inputs/ stages external non-session material → moves to archive/ once processed. drafts/ holds active production (email drafts, agendas, in-progress) → moves to ready/ when finished or archive/ if abandoned. ready/ holds finished work waiting to send → moves to archive/ when sent. archive/ is the final state for retired, consumed, or abandoned work.

Note: reference/core/ holds four identity files (soul.md, offer.md, audience.md, voice.md) — created empty in Mode 1, populated in Mode 2. kits/, frameworks/, brand/, skills/, and qc-agents/ are intentionally empty — they get populated during Mode 3. The projects/ directory starts with only non-project-specific/ — project subdirectories are scaffolded during Mode 2 when an initiative is identified.

Step 2: Create settings.local.json

Create .claude/settings.local.json with exactly this content (identical across all clients — includes both Windows and Linux paths for cross-machine compatibility):

{
  "permissions": {
    "additionalDirectories": [
      "/c/Users/Home/Documents/GitHub/advisory-os-vault",
      "/c/Users/Home/Documents/GitHub/business-aos",
      "/home/kathryn/advisory-os-vault",
      "/home/kathryn/business-aos"
    ]
  }
}

Step 3: Create .gitignore

Create .gitignore with exactly this content (identical across all clients):

node_modules/
.DS_Store
*.tmp

Step 4: Create CLAUDE.md

Create CLAUDE.md at the repo root using the template below. Replace all {{PLACEHOLDERS}} with Mode 1 inputs.

CLAUDE.md Template

# Client Workspace — {{PREFERRED_NAME}} {{LAST_NAME}} / {{COMPANY_NAME}}

**Engagement:** Advisory OS — {{ENGAGEMENT_TYPE}}
**Cadence:** {{CADENCE}}
**Rate:** {{RATE}}

---

## Active Projects

*No active projects yet. Projects are scaffolded when an initiative is identified (Mode 2).*

---

## Session Opener (Every Session — Before Any Work)

1. Ask: "Did you drop anything into the vault-inbox or this project's inbox?"
2. Review last session's state — confirm what moved (sent / used / dead)
3. Execute file moves before starting new work

**Kanban flow:**
- `sessions/` — permanent record, never moves
- `inputs/` → `archive/` once processed (trigger: extracted and incorporated into a deliverable)
- `drafts/` → `ready/` when finished · → `archive/` if abandoned
- `ready/` → `archive/` when sent or published

---

## Weekly Workflow

Post-session processing uses the **9-step cascade** in `.claude/skills/`. Run `/recap-email` to start — each step auto-advances to the next with a human review gate between each.

| Step | Skill | Input | Output |
|------|-------|-------|--------|
| 0 | `/start` | vault-inbox + inbox check | Orientation, file moves |
| 1 | `/recap-email` | Transcript | Gmail draft (client recap) |
| 2 | `/recap-internal` | Transcript | `sessions/recap-YYYY-MM-DD.md` |
| 3 | `/recap-json` | Transcript | `sessions/recap-YYYY-MM-DD.json` |
| 4 | `/update-cpm` | JSON + prior CPM | `constraints/{{PREFERRED_NAME_LOWER}}-constraint-priority-matrix-YYYY-MM-DD.md` |
| 5 | `/update-masterplan` | JSON + master plan | `master-plan/client-master-plan-{{PREFERRED_NAME_LOWER}}-{{LAST_NAME_LOWER}}.html` (in place) |
| 6 | `/update-projectplan` | JSON + project plan(s) | `projects/[active]/project-plan/` (in place) |
| 7 | `/update-blueprint` | JSON + blueprint(s) | `projects/[active]/blueprint/` (in place) |
| 8 | `/draft-actions` | JSON + skills/kits | `drafts/` |
| 9 | `/session-agenda` | All living docs | `drafts/` + Gmail draft (advisory email) |

**Human review gate after every step.** Do not proceed until Kathryn approves.

Full step-by-step documentation: `WORKFLOW.md`

---

## Where Things Are

| What | Where |
|------|-------|
| Client reference data (names, tools, proper nouns) | `reference-data/{{PREFERRED_NAME_LOWER}}-reference-data.md` |
| Constraint briefs ({{PREFERRED_NAME}}'s input) | `constraints/{{PREFERRED_NAME_LOWER}}-constraint-brief-*.md` |
| Constraint priority matrices (dated outputs) | `constraints/{{PREFERRED_NAME_LOWER}}-constraint-priority-matrix-*.md` |
| Master plan (living doc) | `master-plan/client-master-plan-{{PREFERRED_NAME_LOWER}}-{{LAST_NAME_LOWER}}.html` |
| Non-project builds | `projects/non-project-specific/` |
| Session raw material (transcripts, recaps, JSON) | `sessions/` |
| External non-session material (client docs, status updates) | `inputs/` |
| Active production (email drafts, agendas, in-progress) | `drafts/` |
| Finished, waiting to send or publish | `ready/` |
| Retired, consumed, or abandoned | `archive/` |
| Client portrait (soul, offer, audience, voice) | `reference/core/` |
| Client brand kit + skill + QC | `brand/` |
| Client deployment kits (multi-file production systems) | `kits/` |
| Client IP frameworks (custom models, methodologies) | `frameworks/` |
| Client-specific single-file build specs | `skills/` |
| Client-specific QC agents (standalone) | `qc-agents/` |

---

## Connected Repos (via .claude/settings.local.json)

- **advisory-os-vault** — framework kits (CPM, master plan, deployment cycle), production skills, QC agents
- **business-aos** — business context, audience, voice, pricing, decisions

### Vault Frameworks Used

| Framework | Vault Location | Used In |
|-----------|---------------|---------|
| Constraint Priority Matrix | `content/frameworks/kit-constraint-priority-matrix/` | Step 1 |
| Master Plan | `content/frameworks/kit-master-plan/` | Step 2 |
| Constraint Identifier | `content/frameworks/kit-constraint-identifier/` | New constraint extraction |
| Deployment Cycle | `content/frameworks/kit-deployment-cycle/` | Build design |
| Client Email | `content/frameworks/kit-client-email/` | Client communication |
| Session Recap | `content/frameworks/kit-session-recap/` | Post-session emails |

---

## Rules

- Reference data is canonical — overrides transcript misspellings
- Blueprints are {{PREFERRED_NAME}}'s documents — only update for things that change what {{PRONOUN_CAPS}} sees
- Project plans are Kathryn's documents — internal execution detail
- Master plan is Kathryn's strategic command center — never shared with client
- Never update CPM from constraint brief alone — wait for session transcript
- Human review gate after every step — no exceptions

Placeholder Reference:

Step 5: Create WORKFLOW.md

Create WORKFLOW.md at the repo root using the template below.

WORKFLOW.md Template

# Weekly Workflow — aos-client-{{SLUG}}

This is the repeatable process for processing {{PREFERRED_NAME}}'s weekly inputs and updating all living documents.

---

## Weekly Rhythm

| Day | What Happens |
|-----|-------------|
| {{SESSION_DAY}} | Session with {{PREFERRED_NAME}} → transcript exported → Relay produces JSON + recap |
| After session | **In Claude Code:** Run `/recap-email` to start the 9-step cascade |

---

## Before You Start (Every Processing Session)

In Claude Code:

1. **Read `reference-data/{{PREFERRED_NAME_LOWER}}-reference-data.md` first. No exceptions.** Every agent, every session, every document. All proper nouns, names, and tool names in every output must match this file — not the transcript. Do not process anything before reading this.
2. Say: **"Pull the latest for {{PREFERRED_NAME}}"** — Claude syncs the client repo from GitHub
3. Go to Google Drive and grab:
   - Latest session transcript
   - Latest JSON analysis file
   - Latest recap email (produced by Relay)
   - Latest constraint brief (if not already in `constraints/`)
4. Drop transcripts + JSON into `sessions/`. Drop other inputs into `inputs/`.
5. **QC the recap email before anything else ships** — run both checklists (see Email QC step below)
6. Say: **`/recap-email`** — Claude runs the 9-step cascade (Steps 1–9 below), auto-advancing with human review gates

**Before asking Kathryn any status question:** Check the master plan, the most recent CPM, and the session agenda first. If the answer is in those documents, do not ask.

**Never use background agents for document production.** The main conversation has full access to all repos. Background agents are subprocesses — they cannot write to client repos, which forces workarounds that introduce encoding corruption. Use agents only for research or exploration where the result is information returned to the main conversation. All builds, SOPs, CPMs, master plan updates, and blueprint updates must be done in the main conversation.

---

## Email QC

**Triggered by:** Relay-produced recap email in `inputs/`

**Run both checklists in order:**
1. `content/frameworks/kit-session-recap/qc-checklist-session-recap-email.md` — structure, voice, action items, AI tells
2. `content/frameworks/kit-client-email/qc-checklist-advisory-client-email.md` — strategic value, asset completeness, reference data verification, voice

**Reference data check is embedded in checklist 2** — every name and tool in the email must match `reference-data/{{PREFERRED_NAME_LOWER}}-reference-data.md`.

**Fix all fails before proceeding.** The email does not ship and the weekly update does not run until both checklists pass.

---

## The 9-Step Post-Session Cascade

The cascade runs via skills in `.claude/skills/`. Run `/recap-email` to start. Each step auto-advances to the next. Human review gate between every step.

/recap-email (Step 1) → /recap-internal (Step 2) → /recap-json (Step 3) → /update-cpm (Step 4) → /update-masterplan (Step 5) → /update-projectplan (Step 6) → /update-blueprint (Step 7) → /draft-actions (Step 8) → /session-agenda (Step 9)

### Steps 1–3: Session Processing
| Step | Skill | Reads | Produces |
|------|-------|-------|----------|
| 1 | `/recap-email` | Transcript + reference data | Gmail draft (client recap) |
| 2 | `/recap-internal` | Transcript + reference data + prior JSON | `sessions/recap-YYYY-MM-DD.md` |
| 3 | `/recap-json` | Transcript + reference data + prior JSON | `sessions/recap-YYYY-MM-DD.json` |

### Steps 4–7: Living Document Updates
| Step | Skill | Reads | Produces |
|------|-------|-------|----------|
| 4 | `/update-cpm` | JSON + prior CPM + constraint briefs | `constraints/{{PREFERRED_NAME_LOWER}}-constraint-priority-matrix-YYYY-MM-DD.md` |
| 5 | `/update-masterplan` | JSON + master plan | Updated master plan HTML (in place) |
| 6 | `/update-projectplan` | JSON + project plan(s) | Updated project plan HTML(s) (in place) |
| 7 | `/update-blueprint` | JSON + project plan + blueprint(s) | Updated blueprint HTML(s) (in place) |

### Steps 8–9: Action + Prep
| Step | Skill | Reads | Produces |
|------|-------|-------|----------|
| 8 | `/draft-actions` | JSON + skills/kits | Drafts in `drafts/` + summary |
| 9 | `/session-agenda` | All living docs | Session agenda + Gmail draft (advisory email) |

**Step 4 critical rule:** Never update the CPM from the constraint brief alone. Always wait for the session transcript.

**Step 7 critical rule:** Blueprint is the client's document. No internal jargon, no build numbers, no pressure framing.

**After all 9 steps are reviewed and approved, push to GitHub.**

### STOP — Human Review (after every step)

---

## Monthly: Master Plan Comprehensive Review

- Review all constraint statuses for accuracy
- Update recurring themes if evidence has shifted
- Review recommendations pipeline
- Check GPS trajectory against month-over-month progress
- Update `reference-data/{{PREFERRED_NAME_LOWER}}-reference-data.md` if team, tools, or engagement details changed

Additional WORKFLOW.md Placeholders:

If the client has stakeholder sessions: Add additional rows to the Weekly Rhythm table and add a note below it explaining the stakeholder session (who, what day, purpose). See aos-client-rc/WORKFLOW.md for the pattern (Pooja/Thursday).

Step 6: Create README.md

Create README.md at the repo root using the template below.

README.md Template

# Client Workspace — aos-client-{{SLUG}}

**Client:** {{PREFERRED_NAME}} {{LAST_NAME}} / {{COMPANY_NAME}}
**Engagement:** Advisory OS — {{ENGAGEMENT_TYPE}}
**Cadence:** {{CADENCE}}
**Advisor:** Kathryn Brown
**Rate:** {{RATE}}

---

## Folder Structure

` ` `
aos-client-{{SLUG}}/
├── sessions/            ← Permanent record — transcripts, JSON, recaps (never moves)
├── inputs/              ← External non-session material → archive/ once processed
├── drafts/              ← Active production (email drafts, agendas, in-progress)
├── ready/               ← Finished, waiting to send or publish
├── archive/             ← Retired, consumed, or abandoned
├── constraints/         ← Constraint briefs (from {{PREFERRED_NAME}}) + CPMs (produced here)
├── master-plan/         ← Client Master Plan (living doc, updated weekly + monthly)
├── projects/
│   └── non-project-specific/
│       └── ...              ← Builds not tied to a single project
├── reference/
│   └── core/            ← Client portrait — soul, offer, audience, voice
├── reference-data/      ← Canonical source for names, tools, proper nouns
├── kits/                ← Multi-file deployment kits (one subfolder per deliverable type)
├── frameworks/          ← Client IP — custom models, methodologies, scoring systems
├── brand/               ← Client brand kit, brand skill, brand QC
├── skills/              ← Single-file build specs (one file per task type)
├── qc-agents/           ← Standalone QC files (for deliverables without a full kit)
├── WORKFLOW.md          ← Step-by-step processing instructions
└── README.md            ← This file
` ` `

---

## GitHub

| Field | Value |
|-------|-------|
| Repo | `kathryn-art/aos-client-{{SLUG}}` (private) |
| Visibility | Private — only Kathryn |

Note: The folder structure in the README starts minimal. As projects are scaffolded (Mode 2), update the README to reflect the actual project directories. See aos-client-rc/README.md for a fully expanded example.

Step 7: Create Reference Data

Copy the vault template and pre-fill known fields:

1. Copy content/frameworks/kit-reference-data/client-reference-data-template.md to reference-data/{{PREFERREDNAMELOWER}}-reference-data.md
2. Fill in the header: ## {{PREFERREDNAME}} {{LASTNAME}} — {{COMPANY_NAME}}
3. Fill in the Engagement section with known values: entry point, current engagement, rate, start date, meeting cadence, advisor (Kathryn Brown), brand (Advisory OS)
4. Leave Team Roster, Tools & Software, and Key Proper Nouns empty — these get populated during Mode 2

Step 8: Create /start Skill

Create .claude/skills/start/SKILL.md. This is the session opener — syncs the repo, reads reference data, checks for new inputs, shows workflow position, and routes to the next action.

Model on: aos-client-jb/.claude/skills/start/SKILL.md or aos-client-rc/.claude/skills/start/SKILL.md

Adapt for this client:

• Replace repo path with $HOME/aos-client-{{SLUG}}
• Replace reference data path with reference-data/{{PREFERREDNAMELOWER}}-reference-data.md
• Replace client name throughout
• Replace CPM path prefix with {{PREFERREDNAMELOWER}}-constraint-priority-matrix-
• Replace master plan path with client-master-plan-{{PREFERREDNAMELOWER}}-{{LASTNAMELOWER}}.html
• Set git pull origin master (not main)
• Set session day(s) for "next session" reference

Step 9: Create 9-Step Cascade Skills

Create all 9 cascade skill files in .claude/skills/. Each skill gets its own directory with a SKILL.md inside.

Model on: aos-client-jb/.claude/skills/ — these are the tested, working cascade skills. Adapt each for this client by replacing:

• Client name (preferred name + full name + company)
• Reference data path
• File path prefixes (preferred name lowercase)
• Master plan file path
• CPM file path prefix
• Pronoun references

The 9 cascade skills (in auto-advance order):

1. recap-email/SKILL.md — Step 1

• Reads newest transcript in sessions/ + reference data
• Produces Gmail draft (client recap email) to kathryn@creatingyourplan.com
• Engagement work filter: include deliverables/decisions, exclude backstory/venting/personal
• One line per topic in What We Covered, no build numbers
• Auto-advances to /recap-internal

2. recap-internal/SKILL.md — Step 2

• Reads transcript + reference data + prior JSON
• Produces sessions/recap-[YYYY-MM-DD].md
• Full advisor record — capture EVERYTHING including org context
• Auto-advances to /recap-json

3. recap-json/SKILL.md — Step 3

• Reads transcript + reference data + prior JSON
• Produces sessions/recap-[YYYY-MM-DD].json
• Fields: conversation, wins, actions, priorActions, quotes, gpsSignals, seeds, constraintMentions, recommendations, advisorNotes, notCovered
• Use constraint emoji terminology: 🔁 Recurring, 📈 Escalating, 📦 Same Root Cause, ↔️ Solve Together, ⬆️/⬇️ Upstream/Downstream
• Auto-advances to /update-cpm. Note: from this point forward, all steps read the JSON — not the transcript.

4. update-cpm/SKILL.md — Step 4

• Reads most recent JSON + most recent CPM + any newer constraint briefs
• Follows vault CPM framework (content/frameworks/kit-constraint-priority-matrix/)
• Never update CPM from constraint brief alone — wait for transcript
• QC: 90+ to pass
• Produces constraints/{{PREFERREDNAMELOWER}}-constraint-priority-matrix-[YYYY-MM-DD].md
• Auto-advances to /update-masterplan

5. update-masterplan/SKILL.md — Step 5

• Reads most recent JSON + master plan HTML
• Follows vault master plan framework (content/frameworks/kit-master-plan/)
• Process: conversation recap → action items → GPS → constraint statuses → wins → recommendations → history
• Never edit historical recaps or GPS history. Never fabricate quotes.
• Updates master-plan/client-master-plan-{{PREFERREDNAMELOWER}}-{{LASTNAMELOWER}}.html in place
• Auto-advances to /update-projectplan

6. update-projectplan/SKILL.md — Step 6

• Reads most recent JSON + active project plan(s)
• Follows vault project plan framework (content/frameworks/kit-project-plan/)
• Updates GPS (initiative-scoped), build cards, constraint themes, stakeholders, prework, evidence, actions, timeline
• Cross-document consistency with blueprint required
• Auto-advances to /update-blueprint

7. update-blueprint/SKILL.md — Step 7

• Reads most recent JSON + updated project plan(s) + active blueprint(s)
• Follows vault blueprint framework (content/frameworks/kit-blueprint/)
• CRITICAL: Client-facing document. No internal jargon, no build numbers, no pressure framing
• Client-facing status: Green = "On Track", Amber = "Delayed", Red = "On Hold", Gray = "Waiting"
• Cross-document consistency with project plan required
• Auto-advances to /draft-actions

8. draft-actions/SKILL.md — Step 8

• Reads most recent JSON, extracts Kathryn's action items
• Scans skills/, kits/, .claude/skills/ for matching skills/kits
• Match found → run skill → draft to drafts/. No match → spec/brief in drafts/. Not draftable → skip.
• Produces drafts/action-drafts-summary-[YYYY-MM-DD].md
• Auto-advances to /session-agenda

9. session-agenda/SKILL.md — Step 9 (final)

• Reads most recent JSON + CPM + master plan + project plan + blueprint
• Produces two outputs:

1. Session agenda (Kathryn's intelligence brief) → drafts/session-agenda-[YYYY-MM-DD].md
2. Advisory email to client → Gmail draft to kathryn@creatingyourplan.com

• Reports: "Cascade complete. All 9 steps finished."

Every skill includes:

• name: and description: in YAML frontmatter
• Client name and engagement context
• Reference data path and correction rules
• Push instructions (file path, branch: master, commit message)
• Auto-advance instruction (except session-agenda which reports cascade complete)

Step 10: Run Mode 1 Quality Gate

Run the Mode 1 checklist from 04-new-client-kit-quality.md. All 28 items must pass.

Step 11: Initial Commit

git add -A
git commit -m "Day 0 scaffold — {{PREFERRED_NAME}} {{LAST_NAME}} / {{COMPANY_NAME}}"
git push -u origin main

Mode 2: Session 1 Population

Prerequisites

• Mode 1 complete (all 20 quality gate items passed)
• At least one session transcript available
• JSON analysis file available

Step 1: Populate Reference Data

Read the session transcript and JSON analysis from sessions/. Update reference-data/{{PREFERREDNAMELOWER}}-reference-data.md:

1. Team Roster — add every person mentioned by name. Flag transcript misspellings in the Notes column.
2. Tools & Software — add every tool or software mentioned. Note transcript misspellings.
3. Key Proper Nouns — catch anything else that gets misspelled in audio-to-text.
4. Company section — fill in industry, location/time zone if mentioned.
5. Update Log — add entry for initial population.

Step 2: Run First CPM

Framework: Vault content/frameworks/kit-constraint-priority-matrix/ — Mode 1 (no master plan exists yet).

Feed all available files:

1. Reference data (just populated)
2. Client's stated vision/goal (if known)
3. All available constraint briefs (chronological)
4. All available session transcripts (chronological)
5. All available JSON analysis files
6. Any group session data where this client participates

Output: constraints/{{PREFERREDNAMELOWER}}-constraint-priority-matrix-YYYY-MM-DD.md

Quality gate: CPM must score 90+ on the vault 04 checklist.

STOP — Human Review

Kathryn reviews the first CPM. This is the foundation for everything downstream.

Step 3: Create Master Plan

Framework: Vault content/frameworks/kit-master-plan/

Feed the approved CPM + transcript(s) + JSON + reference data.

Output: master-plan/client-master-plan-{{PREFERREDNAMELOWER}}-{{LASTNAMELOWER}}.html

Quality gate: Master plan must pass vault QC (06 file).

STOP — Human Review

Step 4: Scaffold Project (if initiative identified)

If the CPM or session content identifies a clear initiative, scaffold the project directory:

projects/{{PROJECT_SLUG}}/
├── blueprint/
├── project-plan/
├── builds/
├── deployment/
├── drafts/
├── inputs/
└── outputs/

Then produce the project plan and blueprint using vault frameworks:

• Project Plan: Vault content/frameworks/kit-project-plan/
• Blueprint: Vault content/frameworks/kit-blueprint/

STOP — Human Review (for each document produced)

Step 5: Update CLAUDE.md

1. Update Active Projects section with any scaffolded projects (replace the "no active projects" placeholder).
2. Update Where Things Are table with project-specific paths.
3. Update Weekly Workflow table Step 3/4 outputs if project paths are now known.

Step 6: Run Mode 2 Quality Gate

Run the Mode 2 checklist from 04-new-client-kit-quality.md. All applicable items must pass.

Step 7: Run Readiness Checklist

Run the "Ready for Weekly Operations" checklist from 04-new-client-kit-quality.md. All 12 items must pass before the first regular processing session.

Mode 3: Maturity Upgrade

Mode 3 actions are independent. Each section below stands alone. Run whichever action is appropriate based on what patterns have emerged.

Create Brand Kit

When: Brand patterns are clear — typically after 3-5 sessions. The client has preferences for how documents look, specific terminology they use, or you've produced enough builds to see consistent patterns.

What to model on: aos-client-rc/brand/ — Crulliance brand skill, brand QC. Study the structure, then adapt for the new client.

Produce 3 files in brand/:

1. Brand HTML reference ({{BRANDNAMELOWER}}-brand-reference.html) — Visual guide: colors, typography, document themes, usage rules.

1. Brand skill ({{BRANDNAMELOWER}}-brand-skill.md) — Production rules covering:

• Color system (one color = one meaning, consistent across all documents)
• Typography rules (which fonts for headers vs body)
• Terminology rules (what the client calls things)
• Document-specific rules (light theme vs dark theme, which docs get which treatment)

1. Brand QC ({{BRANDNAMELOWER}}-brand-qc.md) — Quality gate for brand compliance. Layers on vault copy QC — adds brand-specific checks, doesn't replace general quality checks.

After creating: Update CLAUDE.md, WORKFLOW.md, and cascade skills:

• Add brand skill as prerequisite in WORKFLOW.md
• Add brand QC to QC Quick Reference
• Add brand files to Where Things Are table if not already listed
• Update cascade skills in .claude/skills/ — Steps 6 (update-projectplan), 7 (update-blueprint), and 8 (draft-actions) should reference the new brand skill for client-facing outputs

Create Client-Specific Skill

When: A document type recurs and needs client-specific production rules beyond what the vault framework provides.

Two types — know which one you're building:

Single-file skill (goes in skills/):

Key rule: The skill references the vault framework — it does not duplicate it. A client-specific master plan skill adds client-specific processing rules on top of the vault master plan framework. It doesn't rewrite the framework.

Naming pattern: {{BRANDNAMELOWER}}-{{DOC_TYPE}}-skill.md

Common single-file skills (in creation order from RC):

1. Master plan skill — client-specific processing order and section rules
2. Project plan skill — document spec for this client's project plans
3. Project plan update skill — rules for updating project plans after sessions
4. Blueprint skill — document spec for this client's blueprints
5. Blueprint update skill — rules for updating blueprints after sessions
6. Session agenda skill — rules for producing session agendas

After creating: Update CLAUDE.md Skills Quick Reference table, WORKFLOW.md step references, and the corresponding cascade skill in .claude/skills/ to reference the new client-specific skill instead of (or in addition to) the vault framework.

Multi-file deployment kit (goes in kits/[doc-type]/):

Key rule: The kit extends a vault framework kit — it does not replace it. File numbering parallels the vault framework (00 start-here, 01 client context, 04 QC extension, 05 build skill). Start-here orients the builder to use the vault kit first, then the deployment kit.

Naming pattern: kits/{{DOC_TYPE}}/ subfolder containing all kit files

Common kits: SOP kit is the first and primary example (kits/sop/). A kit is warranted when the deliverable type has complex quality gates, an extraction methodology, and client-specific brand/visual standards that add up to more than one file can cover.

After creating: Update CLAUDE.md Where Things Are table and WORKFLOW.md references.

Create Client-Specific QC Agent

When: Quality patterns specific to this client emerge — recurring quality issues, client-specific standards, or document types that need targeted quality checks.

What to model on: aos-client-rc/qc-agents/ — study whichever QC matches the document type.

Key rule: The QC agent layers on vault QC — it adds client-specific checks, doesn't replace vault checks.

Naming pattern: {{BRANDNAMELOWER}}-{{DOC_TYPE}}-qc.md

After creating: Update CLAUDE.md QC Quick Reference table, WORKFLOW.md quality gate references, and the corresponding cascade skill in .claude/skills/ to reference the QC agent (e.g., add "Run qc-agents/[name]-qc.md before delivering" to the relevant cascade step).

Mode 3 Placeholder Reference

Template Customization Notes

What Varies Per Client (customize these)

• Client name, company, engagement details in all headers
• File path prefixes (preferred name in lowercase)
• Session day(s) in WORKFLOW.md rhythm table
• Stakeholder sessions (add rows to rhythm table if applicable)
• Pronoun in Rules section and WORKFLOW.md ("HE sees" vs "SHE sees")
• Number of processing days in WORKFLOW.md

What Is Fixed Across All Clients (do not customize)

• 9-step post-session cascade in .claude/skills/ (auto-advance chain, human review gates)
• /start skill structure (sync, reference data, orient, route)
• settings.local.json content (both Windows and Linux paths)
• .gitignore content
• Directory structure (all directories present from Day 0, including .claude/skills/ subdirectories)
• Connected repos (always vault + business-aos)
• Vault Frameworks Used table (same 6 frameworks)
• 5 standard rules in CLAUDE.md
• Monthly review section in WORKFLOW.md
• "Before You Start" section in WORKFLOW.md (pull, grab from Drive, drop into inputs)
• Cascade auto-advance order: recap-email → recap-internal → recap-json → update-cpm → update-masterplan → update-projectplan → update-blueprint → draft-actions → session-agenda
• Gmail draft destination: kathryn@creatingyourplan.com
• Git branch: master (not main)