---
name: journey-mapper
description: Scans a codebase and generates a self-contained HTML service-design journey map — NN/g combined customer journey + service blueprint format. Use when the user wants to map user journeys, create journey maps, build service blueprints, or understand user flows from code. Prompts for codebase path if not supplied. Works with any tech stack.
---

# Journey Mapper

Reads a codebase, thinks like a service designer, and writes a single browser-ready HTML file with all inferred user journeys and service blueprints.

See `REFERENCE.md` in this skill's base directory for the JSON schema, NN/g methodology notes, and emotional arc heuristics.

## Quick start

If the codebase is not already in scope, ask:
> "Which directory should I scan? And where should I save the output HTML?"

## Workflow

**1. Gather context** — ask for anything not already in scope:

| Input | Default |
|---|---|
| Codebase path | Required — ask if missing |
| Output path | `journey-map.html` in codebase root |
| Product name | Used in the HTML title and rail heading |
| Extra context | Design docs, research, README, API specs |
| Scope | Full scan, or specific subdirectory for large codebases |

**2. Scan the codebase** — dispatch Explore, or use Grep + Read for targeted sweeps. Extract:
- Routes, screens, page components, entry points
- Auth roles, user types, permission models, tenant structures
- Onboarding flows, auth gates, redirect chains, feature flags
- API endpoints, background jobs, third-party integrations
- Email templates, push notifications, in-app alerts, webhook payloads
- Error states, empty states, loading patterns, blocked or warning states

For large codebases: routes first → components → API layer → notifications.

**3. Organise into NN/g structure** — think as a service designer:
- **2–5 actors** — who uses the system (infer from auth roles, user types, API consumers)
- **3–6 categories** — thematic groups (Onboarding, Core workflow, Admin, Recovery, Alternative paths…)
- **4–12 journeys** — one per meaningful end-to-end scenario; cover golden paths first, then error/recovery
- Each journey → **2–5 stages** → **2–4 moments** per stage

Per-moment fields to fill — every inferred value gets `[Assumption]` prefix:

| Field | Notes |
|---|---|
| `doing` | What the user physically does. Factual. No `[Assumption]`. |
| `frontstage` | What they see: UI, email, native prompt. Factual. |
| `backstage` | What the code does behind the scenes. Factual. |
| `support` | Which system, service, or API underpins this. Factual. |
| `thinking` | `[Assumption]` — inferred user thought at this moment. |
| `feeling` | Integer 1–5 (1 = very frustrated, 5 = delighted). Use 2–3 for friction-heavy moments. |
| `pain` | `[Assumption]` — inferred friction point. |
| `opportunity` | `[Assumption]` — inferred improvement idea. |
| `evidence` | Leave blank `""` — the human fills this from user research. |

See `REFERENCE.md` for emotional arc heuristics and a worked example.

**4. Build the HTML with `build.py`** (in this skill's base directory, shown above when the skill loads). It validates the JSON, fills in the template, and writes the map. Do not retype `TEMPLATE.html` yourself.
1. Write the JSON to a temporary file, e.g. `journeys.json` in your scratch or temp directory — not in the user's codebase.
2. Run:
   ```bash
   python3 <skill-dir>/build.py journeys.json <output-path> --product "<Product name>"
   ```
3. Exit code 1 means the JSON has errors; the script lists each one and writes nothing. Fix them and run again. Warnings (missing `[Assumption]` prefixes, unused actors, counts outside the structural guidance) don't block the build, but fix the ones that matter.
4. Keep the product name stable across regenerations of the same product — the page title keys the browser's saved annotations.

**If Python 3 isn't available**, check and build by hand instead:
- Every `journey.actor` and `journey.category` matches an `actors[].id` / `categories[].id`; every actor appears in at least one journey.
- All `id` values are kebab-case and unique; `moment.id` values are unique across the whole file. `feeling` is an integer 1–5, or omitted.
- Copy `TEMPLATE.html` to the output path (e.g. with `cp`), then edit three spots in the copy: `<title>Journey Maps</title>` → `<title>[Product name] — Journey Maps</title>`, `<h1>Journey Maps</h1>` → `<h1>[Product name]</h1>`, and the contents of the `<script type="application/json" id="journeys-data">…</script>` block → your JSON. Change nothing else.
- In that JSON, escape every `</` inside string values as `<\/`, or text such as `</script>` ends the data block early and the page renders blank.

**5. Report back:**

- Path the file was saved to
- Journey count · moment count · category count (`build.py` prints these)
- Coverage gaps: flows or actor types with thin codebase evidence
- 2–3 standout pain points inferred from error states, friction-heavy flows, or complex conditionals