03 — GOLDEN EXAMPLE: Kit Builder
The golden example for the Kit Builder is the full population of existing vault kits. This file annotates patterns observed across all of them — what's universal, what varies, and what each kit type does well.
Reference Kits by Type
Standard 6-File Kits
| Kit | Files | QC Format | Golden Example Format | What It Does Well |
|---|---|---|---|---|
| Blueprint | 6 | 100-point, 90 threshold | HTML (fully populated) | Content filtering (in vs. out), forbidden terms, component-level templates |
| Project Plan | 6 | 100-point, 90 threshold | HTML (fully populated) | Status progressions, 10-step ordered update sequence, deploy chain terminology |
| New Client Kit | 6 | Multi-phase gate (Mode 1/2/3) | Structural reference (points to live repos) | Mode-specific inputs, gold standard references to real repos, coordinator pattern |
Extended Kits (7-9 Files)
| Kit | Files | Extra Files | QC Format | What It Does Well |
|---|---|---|---|---|
| CPM | 9 | Instructions, separate HTML output skill | 100-point, 90 threshold + reject triggers | Direction Gate, Unknown status, behavioral modifier, conflict resolution matrix |
| Master Plan | 9 | Full-document QC, input manifest | Multi-phase (create + update + content quality) | Immutable history rule, GPS history tracking, cross-document QC |
| Change Communication | 7 | Consultant methodology | Pass/fail with blocking failures | Two-beat conditional model, team filter gate, anti-pattern focus, voice authenticity |
| Recruiting kits (x5) | 9 each | Consultant methodology, process agent, split golden examples | Binary Gate 1 + 100-point Gate 2 | Gap protocol, extraction interview as primary source, two-track model |
Narrative & Marketing Kits
| Kit | Files | QC Format | What It Does Well | |
|---|---|---|---|---|
| AOS Interactive Narrative | 6 | Checklist (no points) + mandatory copy QC pass | HTML (fully populated) | 7-decision intake, mechanism naming, interactive emotional design, CTA versioning |
| Standard Interactive Narrative | 5 | Checklist (no points) | HTML reference | Simplified version of AOS narrative — same patterns, lighter documentation |
| Offer Page | 8 | Interactive HTML (weighted) | HTML (fully populated) | 13-section fixed architecture, pre-build validation gate, external reference file dependencies |
Lightweight Kits (2-4 Files)
| Kit | Files | QC Format | What It Does Well | |
|---|---|---|---|---|
| Client Email | 2 | 8-point checklist with ship criteria | Golden examples in client repo (not in kit) | Menu-based assembly (not rigid template), AI tell scan, comparison check against prior emails |
| Session Recap | 4 | 6-point checklist with ship criteria | Golden examples in client repo | Dual output (email + Relay app extraction), owner taxonomy, date-from-filename rule |
Universal Patterns (Present in Every Kit)
These patterns appear in every kit regardless of type or size. The kit builder must include all of them:
- Start-here defines scope AND anti-scope. Every kit says what it does AND what it does not do.
- Terminology locks words. Every kit has at minimum: terms that are used (with precise definitions) and terms that are forbidden (with why).
- Output skill is standalone-readable. Restates scope and inputs — never says "see file 01." Someone reading only the output skill can produce the deliverable.
- Quality checks are specific and testable. "Zero internal jargon: no constraint, CPM, upstream, downstream, GPS, Mode" — not "Appropriate language."
- Golden example is never a template. If it has
{{PLACEHOLDER}}tags, it belongs in the output skill. File 03 is a finished deliverable or a structural reference to one. - Content rules are numbered. Not prose paragraphs — numbered, enforceable rules that can be checked one by one.
- Delivery/ship checklist exists. Every kit has a final gate before the output is presented — either in the output skill or the quality file.
Patterns That Vary (Kit Builder Must Decide)
These patterns appear in some kits but not all. The kit builder must determine which apply:
| Pattern | When to Include | Examples |
|---|---|---|
| Content filtering (in vs. out) | When the kit transforms one document into another | Blueprint (project plan → client dashboard), CPM (transcripts → diagnostic) |
| Forbidden terms | When the output has a different audience than the source material | Blueprint (no internal jargon), Change Communication (no AI writing patterns) |
| Gap protocol (binary gate before build) | When missing inputs would produce a fundamentally flawed output | All recruiting kits, Offer Page |
| Two-gate QC (binary + weighted) | When some errors are disqualifying and others are graded | Recruiting kits (Gate 1 blocks, Gate 2 scores) |
| Blocking failures | When certain errors cannot be offset by high scores elsewhere | Change Communication, Recruiting kits, Offer Page |
| External QC dependencies | When the kit's output needs voice/copy validation beyond its own checklist | AOS Narrative (copy-qc.md, sentence-editor.md), Offer Page (copy-qc.md), Client Email (AI tell scan) |
| Consultant methodology file | When the kit involves facilitating a conversation, not just producing an artifact | Change Communication, all Recruiting kits |
| Cross-document QC | When the kit's output must stay consistent with upstream/downstream documents | Master Plan (covers Project Plan + Blueprint too) |
| Split golden examples | When distinct production tracks exist | Recruiting kits (03a consultant, 03b agent) |
| CTA versioning | When the same output ships with different calls-to-action over time | AOS Narrative (launch, product, evergreen versions) |
| Comparison check | When the output should be measured against the prior version sent to the same audience | Client Email, Session Recap |
What Makes a Kit "Complete"
A kit is complete when every applicable item is addressed:
Core (every kit):
- [ ] Start-here answers: What is this? Who is it for? What format? What modes? What files? What it does NOT do?
- [ ] Context answers: What inputs? What validation rules? What priority when inputs conflict?
- [ ] Terminology answers: What terms are locked? What terms are forbidden?
- [ ] Golden example exists — fully populated deliverable, structural reference, or explicit placeholder awaiting first deployment
- [ ] Quality gate exists — format matches the kit's needs (points, pass/fail, checklist, or interactive)
- [ ] Output skill is standalone-readable with scope, inputs, content rules, and delivery checklist
- [ ] Self-improvement loop is built in — Mode 2 in start-here, failure modes in quality, change log convention
Extended (if applicable):
- [ ] Consultant methodology covers how to facilitate the human session
- [ ] Gap protocol defines what blocks the build
- [ ] Cross-document QC validates downstream consistency
- [ ] External QC dependencies are documented (which files, when to run them)
- [ ] Split golden examples cover each production track