---
description: Save your current work. Auto-generates a descriptive save message from context.
argument-hint: "[optional message]"
---

# Save

You are saving the writer's current work. Your job is to create a git commit with a writer-friendly message, without exposing any git terminology.

Follow the auto-invoke policy. In the source repository it is documented at `docs/auto-invoke-policy.md`. `/scr:save` does not spawn agents. It owns safe local helpers for `STATE.md`, `CONTEXT.md`, and `HISTORY.log`, then saves `.manuscript/`.

Resolve all variable context surfaces through `docs/surface-resolution-protocol.md` before reading, writing, or regenerating adapted files.

## What to do

1. **Check for `.manuscript/` directory.** If missing: "No manuscript found. Start with `/scr:new-work`."

2. **Check for `.git/` directory.** If missing:
   - Read `.manuscript/config.json` for `developer_mode`.
   - If `developer_mode: false` (writer mode): silently run `git init` in the project root. No output about this.
   - If `developer_mode: true`: tell the writer "Initializing version tracking for this project." then run `git init`.

3. **Check for changes to save.** Run `git status --porcelain .manuscript/`. If nothing has changed:
   - Look at the last entry in STATE.md "Last actions" table for the timestamp.
   - Tell the writer: "Nothing new to save. Your last save was [timestamp from last action]."
   - Stop here.

4. **Read `.manuscript/STATE.md`** to determine context:
   - Current stage (discuss/plan/draft/review/submit)
   - Current unit name and number (from config.json `command_unit` and STATE.md `Current unit`)
   - Last command run (from "Last actions" table)
   - Record highlights from `.manuscript/RECORD.md` when present: open threads, next-unit obligations, and recent established items

5. **Auto-generate the commit message** based on context:
   - If the writer provided an optional message argument: `"Saved: {writer's message}"`
   - After drafting (last command was `draft`): `"Saved after drafting {unit_name} {N}"`
   - After review (last command was `editor-review`): `"Saved after editor review of {unit_name} {N}"`
   - After revision (last command was `revise`): `"Saved after revising {unit_name} {N}"`
   - After planning (last command was `plan`): `"Saved after planning {unit_name} {N}"`
   - After discussion (last command was `discuss`): `"Saved after discussing {unit_name} {N}"`
   - Default / manual save: `"Saved work in progress on {unit_name} {N}"`

6. **Update STATE.md** "Last actions" table with a new row:
   - Timestamp: current date/time
   - Command: `save`
   - Unit: current unit from STATE.md
   - Outcome: the generated message

7. **Regenerate `.manuscript/CONTEXT.md`** before staging. This file is the bootstrap any new session reads first. Use the `templates/CONTEXT.md` scaffold and fill every `{{TOKEN}}`:
   - `{{LAST_UPDATED}}` -- ISO 8601 timestamp of this save
   - `{{LAST_COMMAND}}` -- `/scr:save`
   - `{{TITLE}}`, `{{WORK_TYPE}}` -- from `.manuscript/config.json`
   - `{{PHASE}}`, `{{CURRENT_UNIT}}` -- from STATE.md (the values you computed in step 6)
   - `{{RECENT_ACTIONS}}` -- the last 5 lines of `.manuscript/HISTORY.log` rendered as table rows. If HISTORY.log is missing, use the last 5 entries from STATE.md "Last actions" instead.
   - `{{UNITS_PENDING_REVIEW}}`, `{{UNITS_PENDING_DRAFT}}`, `{{VOICE_WARNINGS}}`, `{{CONTINUITY_FLAGS}}`, `{{SCAFFOLD_MARKERS}}` -- from STATE.md "Pending"
   - `{{RECORD_HIGHLIGHTS}}` -- a compact 3-5 line summary from RECORD.md: open threads, promises needing payoff, and next-unit obligations. If RECORD.md is missing, use `No RECORD.md found yet.`
   - `{{NEXT_STEP}}` -- the same suggestion `/scr:next` would emit
   - `{{LAST_SCAN}}`, `{{LAST_SCAN_VERDICT}}` -- from STATE.md if recorded; otherwise `never run` and `unknown`
   Save to `.manuscript/CONTEXT.md`. This file is committed alongside STATE.md.

8. **Regenerate `.manuscript/PROGRESS.md`** before staging. This is the openable per-unit progress ledger. Use the `templates/PROGRESS.md` scaffold and derive per-unit status from disk per `docs/progress-protocol.md` (plan, draft, and review files reconciled with STATE.md). Fill the unit ledger, the deliverable progress bar, the pipeline position, and the bucket counts (done / in progress / untouched). Save to `.manuscript/PROGRESS.md`; it is committed alongside STATE.md and CONTEXT.md.

8b. **Regenerate the adapted relationship surface for canonical `RELATIONSHIPS.md`** when the work type has a cast surface (per `surface_applicability`) and the adapted cast surface for canonical `CHARACTERS.md` defines two or more characters, concepts, or figures. This derived relationship map is rebuilt from the relationship sections of the adapted cast surface per `docs/relationships-protocol.md`: every pairing accounted for (pairs with no relationship marked `none`, pairs no one has described yet surfaced as undefined). Save it to the adapted relationship filename (`RELATIONSHIPS.md`, `DEPENDENCIES.md`, or `LINEAGES.md` as resolved from `file_adaptations`). Skip silently for work types without cast or with fewer than two entries. It is committed alongside `STATE.md`, `CONTEXT.md`, and `PROGRESS.md`.

8c. **Regenerate `.manuscript/CONFLICTS.md`** when the work has a central conflict in `WORK.md` or two or more cast entries, and the work type is narrative (skip poetry and speech). This derived conflict map is rebuilt from the `WORK.md` central conflict and the adapted cast surface per `docs/conflict-protocol.md`: every pair accounted for (pairs with no conflict marked `no conflict`). Skip silently where conflict does not apply. It is committed alongside the other derived surfaces.

8d. **Regenerate `.manuscript/PEOPLE-DYNAMICS.md`** when the work type has a peoples surface (per `surface_applicability`) and `.manuscript/PEOPLES.md` defines two or more peoples. This derived people-dynamics map is rebuilt from the "Relations with other peoples" sections of `PEOPLES.md` per `docs/people-dynamics-protocol.md`: every pair accounted for (pairs with none marked `no dealings`). Skip silently where peoples do not apply. It is committed alongside the other derived surfaces.

8e. **Regenerate `.manuscript/GEOGRAPHY.md`** when `.manuscript/PLACES.md` exists and the work type has an adapted world surface (per `surface_applicability`). This derived geography map is rebuilt from `PLACES.md` plus the adapted world surface per `docs/world-layers-protocol.md`: place hierarchy, routes, distance logic, borders, and undefined spatial questions. Skip silently where the world surface does not apply or no confirmed places exist. It is committed alongside the other derived surfaces.

9. **Append one line to `.manuscript/HISTORY.log`** per `docs/history-protocol.md`:
   ```
   {ISO timestamp} | scr:save | message="{generated message}" | files={changed file count} | outcome=committed
   ```
   If HISTORY.log does not exist, create it. Do not stage it as a separate operation -- step 9 picks it up.

10. **Execute the save:**
   ```
   git add .manuscript/
   git commit -m "{generated message}"
   ```
   This commit must include the `STATE.md`, `CONTEXT.md`, `PROGRESS.md`, and `HISTORY.log` updates from steps 6 through 9 so the worktree is clean immediately after a successful save.

11. **Tell the writer** the result (see output section below).

## Automation Status

Every response must include a compact status block:

```text
Automation status:
Trigger: /scr:save
Spawned agents:
- none
Candidate agents:
- none
Local operations:
- STATE.md updated: yes/no
- CONTEXT.md regenerated: yes/no
- PROGRESS.md regenerated: yes/no
- RELATIONSHIPS.md / adapted relationship surface regenerated: yes/no
- CONFLICTS.md regenerated: yes/no
- PEOPLE-DYNAMICS.md regenerated: yes/no
- GEOGRAPHY.md regenerated: yes/no
- HISTORY.log appended: yes/no
- manuscript files saved: yes/no
Candidate local helpers:
- /scr:scan if saved state and disk still disagree
Manual gates:
- none
Auto-invoked:
- /scr:next route computed for CONTEXT.md: yes/no
Why: save uses deterministic local bookkeeping, not a spawned agent
```

## Writer mode output

- **Writer mode** (`developer_mode: false`): "Saved. You can see your save history with `/scr:history`."
- **Developer mode** (`developer_mode: true`): Show full git commit output including the short hash and message: "Committed: {hash} - {message}"

## Edge cases

- **No changes to save:** "Nothing new to save. Your last save was [timestamp]."
- **Not in a Scriveno project** (no `.manuscript/` directory): "No manuscript found. Start with `/scr:new-work`."
- **Git repo corrupted or in bad state:** In writer mode, say "Something went wrong saving your work. Try again, or ask for help." In developer mode, show the git error.
- **Very first save** (no previous commits): Auto-generate message as "Initial save of {work title}" using the title from `.manuscript/WORK.md` if available.

## Response Contract

Every writer-facing response must end with one to four next-command suggestions. Each suggestion must include a short explanation of what that path will do.

The final visible section of every writer-facing response must be the `Next commands:` block. This applies to successful completion, partial completion, blocked, stopped, validation-failed, and prerequisite-missing responses. Do not end with only a summary, report, checklist, external action, upload instruction, or prose-only options.

Use the invocation style for the active runtime when writing command suggestions. Source command IDs use `/scr:*`; Claude Code installed commands use `/scr-*`; Codex installed skills use `$scr-*`. Suggest only runnable Scriveno commands that exist in the installed command surface. Do not invent adjacent workflow names.

Use this format:

```markdown
Next commands:
- `/scr:...`: One short sentence explaining what this path will do.
- `/scr:...`: One short sentence explaining what this alternate path will do.
```

If exactly one path is clearly best, provide one suggestion. If two, three, or four useful paths exist, show them as alternatives. Do not force a linear path when the writer has a real choice.

If the writer seems unsure or no specific next command is obvious, include this default option:

```markdown
Next commands:
- `/scr:next`: Inspect the project state and choose the right next step.
```

If the command stops because a prerequisite is missing, suggest the command that fixes the prerequisite. Keep every explanation practical and writer-facing.

## Tone

Brief. Reassuring. The writer should feel that their work is safely stored. Don't explain what git did -- just confirm it's saved.