adopt

Adopt an existing draft into a Holocron project. Sibling of /holocron:new — same six-phase wizard, plus an extraction phase before it (so the wizard's questions arrive pre-filled with everything we could derive from your prose) and a rewrite phase after it (so your existing chapters land in chapters/ rewritten to the project's craft rules, one approval at a time).

Adopt Context

• Pandoc: !pandoc --version 2>/dev/null | head -1 || echo "NOT INSTALLED"
• Input argument: $ARGUMENTS

What this does

Phase	What happens	Output
A — Input & format check	Resolve the path argument; if .docx, run pandoc preflight + convert to markdown; if .md, use directly	working markdown of the draft
B — Project scaffold (Phase 0)	Ask for a project folder name (default: kebab-case slug from draft title or filename); copy the standalone-story template into that folder	Folder structure + .holocron-checkpoint.json with Phase 0 complete
C — Draft import	Copy original to imports/original.{md,docx}; split markdown on chapter headings; write splits to imports/original-split/chapter-NN.md	Preserved original + per-chapter splits
D — Extraction	Dispatch draft-extractor agent. It spawns reader-digest + voice-analyzer in parallel and synthesises the wizard's slot values	.holocron-adoption.json at project root + craft/voice-analysis.md (raw evidence — the author will read it and author craft/voice.md themselves during the wizard's Voice phase) + craft/tone.md (mechanics only — observed POV/tense/style-guide)
1–7 — Wizard (pre-filled)	Dispatch new-project agent. It detects adoption mode (.holocron-adoption.json present), opens each phase with the extracted defaults, and lets you accept / edit / skip exactly as the normal wizard would. Phase 2 (Voice) surfaces craft/voice-analysis.md as evidence and asks the writer to author craft/voice.md themselves	craft/voice.md, CLAUDE.md, world/era.md, architecture.md, themes.md, characters/*.md, world/factions/, world/locations/, world/lore-overlay/, outline.md, roadmap.md, dashboard
E — Rewrite gate	For each imports/original-split/chapter-NN.md, dispatch writer with the original + full project context + a "rewrite, don't draft fresh" directive. Present the result and ask: accept / revise with notes / skip and stop	chapters/NN-<title>.md per approved chapter

The resulting project is indistinguishable from one created by /holocron:new — same files in the same places — but seeded with your existing material instead of from a blank page.

Usage

/holocron:adopt path/to/the-price-of-glory.md
/holocron:adopt path/to/the-price-of-glory.docx
/holocron:adopt                                    # asks for the path interactively

What to do

0. Verify Pandoc Is Installed (if input is .docx)

Check the Pandoc: line in the Adopt Context block above. If the input argument ends in .docx and pandoc reads NOT INSTALLED, stop immediately. Print the cross-platform install instructions (see the Dependencies section in /holocron:build) and ask the writer to install pandoc, then re-run /holocron:adopt. Markdown input does not need pandoc.

If pandoc shows a version (e.g. pandoc 3.1.x) or the input is .md, proceed.

1. Resolve the input path

• If no argument is given, ask the writer: "Where is your existing draft? Give me a path to a .md or .docx file."
• Verify the file exists with Read (it'll error cleanly if not).
• Note the format: markdown if .md, docx if .docx. Reject other formats with a clear "convert to .md or .docx first" message; markdown is the simplest path.

2. Ask for a project folder name

Suggest a default:

• For a .docx, peek at the document title (pandoc can extract it) or use the kebab-case filename stem.
• For a .md, read the first H1 (# Title) or use the kebab-case filename stem.

Use AskUserQuestion to confirm or change the folder name. Then ask for the absolute path to create the project in (default: current working directory).

3. Scaffold the project (Phase 0 of the wizard)

Copy the standalone-story template:

cp -r ${CLAUDE_PLUGIN_ROOT}/templates/standalone-story/ ./<project-name>/

Write .holocron-checkpoint.json in the new project root with lastCompletedPhase: 0 and phases.0_setup.completed: true. This tells the wizard (when we dispatch it later) to begin at Phase 1.

4. Import the original draft

Inside the project:

mkdir -p imports/original-split

Copy the original to imports/original.<ext>:

cp <input-path> ./imports/original.<md|docx>

If the input was .docx, also produce a markdown working copy:

pandoc imports/original.docx -o imports/original.md --wrap=none

(The .docx original is preserved alongside the markdown working copy. We work off the .md from this point on.)

5. Split on chapter headings

Read imports/original.md. Detect chapter boundaries using these rules, in order of preference:

• H1 headings (# Chapter 1, # Chapter One, # Prologue, # Epilogue, # 1, # I.)
• H2 headings (## Chapter 1, etc.) — fall back if no H1 chapter pattern matches
• A bare numeric or roman-numeral line on its own (less reliable; use only if no markdown headings exist)

For each chapter found, write imports/original-split/chapter-NN.md where NN is zero-padded (01, 02, ..., 99). Include the original heading as the file's first line so the writer can see what was matched. If no chapter boundaries are found (one continuous prose dump), ask the writer how to split it before continuing — don't guess.

6. Dispatch the draft-extractor

Task tool with subagent_type: "holocron:draft-extractor"
prompt: "Extract Holocron adoption data from the splits in ./imports/original-split/.
        Read every chapter. Spawn reader-digest (careful mode) and voice-analyzer
        in parallel against the splits. Synthesise the wizard's slot values into
        .holocron-adoption.json at the project root. Write craft/voice-analysis.md
        as raw evidence from the voice-analyzer's report (the author authors
        craft/voice.md themselves during the wizard's Voice phase — DO NOT write
        voice.md). Write craft/tone.md from observed mechanics only (POV, tense,
        style-guide hint, dash style, dialogue quotes) — NOT the voice itself.
        Use the schema documented in your system prompt."

Wait for the extractor's summary. Confirm .holocron-adoption.json, craft/voice-analysis.md, and craft/tone.md were written before moving on. Do not proceed if craft/voice.md already exists at this stage — that file is the author's to author during the wizard, not the extractor's to write.

7. Dispatch the wizard in adoption mode

Task tool with subagent_type: "holocron:new-project"
prompt: "ADOPTION MODE. Phase 0 is already complete — the project folder is
        scaffolded, the original draft is in imports/, and the draft-extractor
        has populated .holocron-adoption.json + craft/voice-analysis.md +
        craft/tone.md (mechanics only).

        Begin at Phase 1. At each phase, read the relevant section of
        .holocron-adoption.json and present its values as the *recommended
        default* using your existing Socratic frame. Honour the confidence
        flags — high-confidence values lead the question, low-confidence
        values are asked openly as if no extraction had happened.

        At Phase 2 (Voice): DO NOT auto-write craft/voice.md. Surface
        craft/voice-analysis.md to the writer as raw evidence, then walk them
        through authoring craft/voice.md in their own words — Path B (Write
        fresh, guided) is the default in adoption mode, but offer Path A
        (Import from another project) if the writer prefers. Generate CLAUDE.md
        from voice.md once written.

        Run through phases 1–7 as normal. Return when the dashboard is installed."

The wizard now drives. You wait for it to finish.

8. Rewrite gate (Phase E)

Once the wizard returns and the project is fully scaffolded with the writer's confirmed answers, enter the rewrite phase.

For each chapter in imports/original-split/ in numeric order:

1. Compute the target filename. Read the chapter's first line for a title; produce chapters/NN-<kebab-title>.md. If no title, use chapters/NN-untitled.md.

2. Dispatch the writer:

   Task tool with subagent_type: "holocron:writer"
   prompt: "REWRITE MODE — this is an adoption flow.
   
   The original chapter is at imports/original-split/chapter-NN.md. Read it.
   Preserve plot beats, character actions, world details, and the AUTHOR's
   declared voice (in craft/voice.md — authored by the writer during the
   wizard's Voice phase from the voice-analyzer's evidence). Apply project
   craft rules: no filter words, no summarised emotion, scene structure
   (Goal → Conflict → Disaster), audiobook awareness, prose rhythm.
   
   Read the standard project context first, in this priority order:
   1. CLAUDE.md (project root — shared agent context, author voice section)
   2. craft/voice.md (the AUTHOR's canonical voice — overrides everything below)
   3. craft/tone.md (mechanics — POV, tense, style guide)
   4. architecture.md
   5. outline.md entry for this chapter
   6. characters/
   7. world/
   8. prior already-rewritten chapters chapters/01..(NN-1).md for voice continuity
   
   Write the rewritten chapter to chapters/NN-<title>.md. Output the prose
   only — no preamble, no commentary."
   

3. Present the result. Show the writer the rewritten chapter (file path + first 500 words or so for in-line preview). Then use AskUserQuestion:

   "Chapter NN rewritten. What's next?"

   • Accept and continue to the next chapter (recommended)
   • Revise this chapter — I'll feed your notes back to the writer
   • Skip and stop here — leave the rest of imports/ for later

4. On Accept: advance to the next chapter.

5. On Revise: ask for notes (free-form), then re-dispatch the writer with the same prompt plus a Revision notes: block. Loop on this same chapter until Accept or Skip.

6. On Skip: record which chapter we stopped at; break the loop; tell the writer how to resume (run /holocron:adopt again on the same project — it'll detect existing chapters/*.md and offer to resume from the next unrewritten chapter).

9. Closing summary

Once the rewrite loop ends (all chapters rewritten or skipped), produce a short summary:

Adoption complete.

  Project:         <project-name>
  Source:          imports/original.<ext>  (preserved)
  Splits:          imports/original-split/  (NN chapters)
  Wizard phases:   1–6 confirmed via /holocron:new pre-fill
  Rewritten:       chapters/01..MM-<title>.md  (MM of NN approved)
  Remaining:       NN - MM chapters still in imports/  (skipped — resume any time)

Next steps:
  /holocron:next        — what to work on now
  /holocron:dashboard   — refresh the dashboard
  /holocron:write next  — keep going from the last rewritten chapter

Resumability

/holocron:adopt writes the same .holocron-checkpoint.json that /holocron:new does. If the writer breaks mid-flow (in extraction, in the wizard, or in the rewrite gate), /holocron:go will load the project and offer to resume at the right phase.

For the rewrite gate specifically: if chapters/01..K-*.md already exist when adopt is re-run on an existing project, it offers to resume from chapter K+1 in imports/original-split/ without re-running extraction or the wizard.

Defaults

If the writer doesn't specify, use these:

• Template: standalone-story (matches /holocron:new's default)
• Style guide: inferred by the extractor from spelling patterns (CMOS for American spellings, NOSM for British) — the wizard can override in Phase 3 (Premise) when it confirms language/style settings
• Era: the extractor's best guess from the prose; low-confidence → wizard asks explicitly in Phase 1

Specialists invoked along the way

• draft-extractor — the extraction coordinator (Phase D); spawns the readers and the voice analyser
• reader-digest — coordinates parallel reader-careful runs across chapters
• reader-careful — per-chapter extraction (plot, characters, themes, craft)
• voice-analyzer — POV / tense / rhythm / distance / characteristic moves → craft/voice-analysis.md (raw evidence; the author authors craft/voice.md from it during the wizard's Voice phase)
• new-project — the wizard (Phases 1–7) in adoption mode, including the Voice phase (Phase 2) where the author authors craft/voice.md themselves
• architect / character-developer / outliner / world-builder / lore-researcher — spawned by new-project as it always does, on phases that need deep work
• writer — the rewriter (Phase E), one chapter at a time, with approval gates

What this does NOT do

• No folder-of-files input mode in v1.1.0. One .md or one .docx only. Writers with a folder should concatenate first.
• No silent era relocation. If the draft is clearly not Old Republic and the extractor flags low confidence on era, Phase 1 asks the writer explicitly — never silently transposes.
• No batch rewrite. Always chapter-by-chapter with an approval gate. Prevents voice drift compounding before the writer sees it.
• No deletion of originals. imports/original.* and imports/original-split/* are preserved forever; rewritten output lives under chapters/ per Holocron convention.