---
name: |-
pick-the-right-document-router
description: |-
Entry point for any 'write up / document / spec / render a doc' request. Reads the ask, picks the matching Playbooks template, decides where it lives, and calls the layered diagram skill only for the levels that doc needs. Use instead of invoking layered diagrams directly when the output is a document. Not for a pure diagram request, quick chat answers, or Second Brain core notes.
notion_page_id: 4169439f-1486-8186-a8d0-fd13cdec6625
---
## When to use
Use when I ask for a document of any kind: "write this up", "spec this", "document X", "render a doc for…". This page is the interface. It picks ONE template, then loads only what that template needs.
Not for:
- A pure diagram ask ("diagram X", "state machine for Y") → go straight to .
- A single idea to remember → Second Brain core note, owned by the Second Brain operator.
- A quick answer I'll read once → stay in chat.
## Step 1 · Route the request
Match the question I'm really asking to one row. Phase follows .
| If I'm asking… | Template | Phase | Diagrams (layered levels) |
| --- | --- | --- | --- |
| "Is this worth building? What's the problem?" | | Discover | None |
| "What did users / competitors say?" | | Discover | None |
| "Start a brand-new product from scratch" | | Discover → Define | None |
| "What are we building, for whom, what's in scope?" | | Define | None (module map as a list) |
| "How does a user move through it? Journey, flow, onboarding" | | Define | User flow: one plain Mermaid flowchart (the template's Flow section). No layered levels |
| "How should this feel / look for the user? Screens, wireframes, onboarding or setup redesign" | | Define → Design | User flow (Mermaid flowchart) + lo-fi wireframes of each journey's happy path in Paper or Figma, made before code and linked from the page. No layered levels |
| "One new feature or one change to an existing module" | | Define | Only the level the change touches |
| "What's the shape of the solution? Boundaries, lifecycles" | | Design | Levels 1 + 2 |
| "Deep dive on one module" (forward or as-built from code) | | Design | Levels 1 + 2 + 3 (full) |
| "Exactly how will it be built? Data model, interfaces, rollout" | | Design | Levels 1 + 3 |
| "Why did we choose X over Y?" | | Design | None |
| "How does this platform work?" (explain an existing tool) | | Any | Level 2 only, if it helps |
| "Something broke, how do I fix it?" | | Launch and run | Level 2 only (how parts connect) |
| A design idea not proven yet | (skill) | Design | As that skill says |
Tie-breakers: one journey as steps → User journey; screens and why they look that way → UX design proposal; what changes in the module and data → Change doc. A UX proposal usually comes first, then feeds the Change doc.
If two rows fit, ask ONE question with the two doc names and a one-line "pick this if…" for each. Don't guess.
If nothing fits, say so, use the nearest template, and log ⚑ "New doc type?" in the Improvement backlog.
## Step 2 · Decide where it lives
- About a specific project → that project's page in Projects.
- A design that isn't built or proven yet → Blueprints, as Draft.
- Generic, reusable knowledge → not a doc; hand off to the Second Brain operator as core notes.
- If I named a place, use it.
## Step 3 · Build it
1. Fetch the chosen template and keep its sections and order. Drop a section only if it's truly empty, and say which.
2. Fill it in plain product language first, technical detail after.
3. For diagrams, load with ONLY the levels from the table, and use one worked example across the whole doc.
4. Unknown facts go under Open questions with ⚑. Never guess.
5. Size the "why" to the bet: a new product or anything over about two weeks or with real spend gets a written benefits statement (one-pager, or the Change doc's benefits table); smaller work gets problem, why now and one goal; a story gets its "so that…".
6. **No repo artifacts.** If the doc is written from a codebase (brownfield PRD, as-built module spec, resync, or a new module built on existing code), read the code but write only what it does, in product words. Link the repo once in the header. Never paste file or folder paths, class, function or component names, table or column names, routes, SQL, schemas, commit hashes or code. Only the Tech spec may carry low-level detail.
## Output
One Notion page from the chosen template, in the place from Step 2. Reply with: the doc type and why (one line), the link, and every ⚑ item.
## Rules
- One request, one doc. If I'm really asking for two (e.g. PRD + tech spec), say so and build the earlier phase first.
- Never invoke the layered skill at full depth by habit; the table decides the levels.
- Don't edit the templates themselves; suggest changes in the Improvement backlog.
## Done checklist
- [ ] The template choice is stated with a one-line reason.
- [ ] Only the diagram levels in the table were produced.
- [ ] The page sits in the right place and links back to its project or Blueprint.
- [ ] Report back: doc type, link, ⚑ items.
## Handoff
Business cases and priorities → CPO. Product docs → Product manager. HLDs, tech specs and ADRs → CTO. UX proposals and user flows → Head of design. Delivery and milestones → Project manager. Blueprints → stays with the Chief of staff until I promote it. Generic learnings → Second Brain operator.