Skip to content
Editorialist 3.5.1

Importing Reviews

This page documents what can come back from a reviewer or AI, where each object goes, and how a review batch gets into your vault.

The formats are written for an AI to produce. For an AI review, paste the formatting instructions into the conversation along with your prose. When the feedback comes from a human — margin notes on a printed page, a marked-up document, an email — hand their notes to an AI together with these same instructions, and it shapes them into Editorialist-ready output with your human reviewer credited as the contributor. A human never works from this format directly.

Tip: you never need to write this format by hand. The review launcher’s Copy instructions for AI button (step 1) puts the full specification — including your book’s real scene IDs, or the names of notes that have none — on the clipboard, ready to paste into an AI conversation.


The launcher’s template includes two output formats. They are different objects with different jobs:

Object Use it for When you get it Where it goes
Review batch Concrete suggestions for specific passages in specific scenes After an AI review, or after an AI converts human notes into Editorialist format Imported through the review launcher, then split into per-scene review blocks
Review block The scene-local copy of the imported suggestions Created by Editorialist when you import a review batch Appended to the bottom of each targeted scene note
Editorialism file Structural or manuscript-level guidance that should be worked as a checklist When feedback is broader than scene-targeted line edits Saved as a separate file under Editorialist/<Book>/<Title>.md

Most revision passes use a review batch. Use an Editorialism file when the output is a durable checklist: subplot work, design rules, multi-scene directives, or manuscript-level guidance.


A review batch is usually a fenced code block labeled editorialist-review. It is the AI response you copy back into Editorialist. Replace attribution placeholders with the actual reviewer: credit the human when converting their feedback and omit Provider/Model; for an independent AI review, use its actual identity without guessing a model version. Keep any additional AI advice separate from the human’s notes.

```editorialist-review
Template: Editorialist advanced
Reviewer: <Actual reviewer name or model name>
ReviewerType: <Accepted role for the actual reviewer>
Provider: <Actual AI provider; omit for human feedback>
Model: <Actual model if known; otherwise omit>
=== MEMO ===
Strengths:
What is working across the scenes you reviewed.
Issues:
Patterns or risks to surface before the author works through the line edits.
=== EDIT ===
SceneId: scn_first_scene_id
Original: ...
Revised: ...
Why: ...
=== CUT ===
SceneId: scn_xxxxxxxx
Target: ...
Why: ...
```

Fences are optional. Most chat UIs strip the outer triple-backtick fence when you copy a reply. The importer accepts both fenced and unfenced output — what matters is the metadata header and the === SECTION === markers. Decorative divider lines some LLMs emit between sections (⸻, ---, ***, ═══) are skipped harmlessly.

When imported, Editorialist groups the batch by target scene and appends one review block to the bottom of each targeted scene note. The block stores the suggestions and memo text for that scene; it does not apply changes to the manuscript.

The lines before the first === SECTION === marker identify the batch and the contributor:

Field Purpose
Reviewer: Display name of the contributor (person or model)
ReviewerType: Role. Human: editor, developmental-editor, line-editor, copy-editor, publisher-editor, agent, beta-reader, sensitivity-reader (human-editor is accepted as editor). AI: ai-editor, ai-developmental-editor, ai-line-editor, ai-copy-editor. Anything unrecognized is treated as the author.
Provider: / Model: For AI contributors — drives the provider brand icon in the contributor directory
Template: / TemplateYear: / SupportedOperations: Emitted by the template; identifies which format version produced the batch

MEMO and QUERY carry commentary rather than a line edit — neither ever applies to the prose. The five operations below them are actionable suggestions (their UI labels are Edit, Move, Cut, Condense, Expand):

Section Fields What it does
=== MEMO === freeform, optional Strengths: / Issues:, optional SceneId: or Note: Commentary that doesn’t belong inline as a line edit. A MEMO with a SceneId or Note attaches to that scene, whether or not the scene received edits; a MEMO without either is duplicated to every scene in the batch. Use as many as needed.
=== QUERY === Id:, optional SceneId: or Note:, Question:, Answer:, optional Recommendation: Answers an author query — a %%query: …%% marker the author left inline. See Author queries below.
=== EDIT === SceneId:, Original:, Revised:, Why: Replace Original text with a specific suggested change.
=== MOVE === SceneId:, Target:, Before: (or After:), Why: Relocate the target passage relative to a destination anchor.
=== CUT === SceneId:, Target:, Why: Remove the target passage. Accepted cuts can be backed up to a cut file first.
=== CONDENSE === SceneId:, Target:, optional Suggestion:, Why: Tighten the passage between two anchors, with a suggested replacement or advisory guidance.
=== EXPAND === SceneId:, Target:, optional Suggestion:, Why: Develop, slow down, or decompress a beat with finished prose or advisory guidance.

Any of the five actionable entries may also carry an optional Context: field — see Context across scenes.

A fix often depends on another scene: the same object, date, or event recurs there, or an earlier scene already established what this one changes. Context: names those passages, one reference per line — the scene number, then a fragment copied verbatim from that scene. A span uses the same "opening" → "closing" form as a CONDENSE target:

=== EDIT ===
SceneId: scn_xxxxxxxx
Original: ...
Revised: ...
Why: Scene 65 already treats the reunion as settled.
Context: 65 "the reunion at Terminus"
- 64 "Wala3 arranges the handover" → "through an XO."

When you review that suggestion, the referenced paragraphs appear in the Across scenes section of the suggestion card, labeled Quoted reference, each with Open beside to read it in a side pane without losing your place. Context is never applied to the manuscript. A line that does not read as a scene number and a quoted fragment is ignored rather than guessed at. Self-contained line edits need no Context:.

A batch does not need any line edits. An editorial letter, a developmental read, or an agent’s notes often arrive as commentary alone, and the natural conversion is one MEMO per scene discussed plus unscoped MEMOs for the manuscript as a whole. Each scoped memo is appended to its own scene as a review block; unscoped memos go to every scene in the batch. When a batch has no edits and no SceneIds or Notes at all, the memos attach to the scene you have open. A memo whose SceneId or Note matches nothing is not imported. The launcher says so before you import — a warning under the paste box, and a line in the destination preview naming the memo — rather than dropping it quietly. A clipboard import that would leave a memo behind opens the destination preview instead of importing straight away; Import placed entries there brings in everything that has a home.

An author query is a %%query: <question>%% marker you leave inline in a scene — run Insert author query (command palette, editor right-click menu, or the panel’s … → Insert author query action) to drop one at the cursor. Obsidian hides it in reading view, and it travels with the manuscript until the next review pass answers it — your editor’s next round, or an AI pass in between.

A query is hidden only inside Obsidian. A manuscript export keeps it unless you strip author queries, so anyone reading that file sees it: keep them in what you send a reviewer, and strip them from anything meant for readers. Radial Timeline’s export has a Strip author queries toggle for this.

Markers written before 3.0.0 use the original name, %%ai: <question>%%. Editorialist still reads, answers and resolves them; new markers are written as %%query:.

The round trip:

  1. Marker. You (or Editorialist’s Insert author query action) write %%query: <question>%% into the scene.
  2. Template. When you copy the formatting instructions, Editorialist finds every marker in the passage, strips it from the copy sent to the reviewer, and lists the questions with instructions to answer each one in its own === QUERY === block.
  3. QUERY answer. The reviewer’s reply includes a === QUERY === section per question: the repeated Question:, a direct Answer:, and an optional one-line Recommendation:.
  4. Resolve or dismiss. On import, the answer routes back to the scene the marker sits in. Resolving a query strips the matching %%query:…%% marker from the note; dismissing it records your decision but leaves the marker in place for a later pass.

The CONDENSE target uses a two-anchor format:

Target: "<verbatim opening fragment>" → "<verbatim closing fragment>"

Both fragments should be copied from the manuscript (≤12 words each is plenty — they’re anchors, not the whole passage). Editorialist tries exact matching first, then a match that tolerates differences in quotes, dashes, spacing, and italics or bold. A paraphrased description still routes the suggestion to “Passage not located” and you can’t act on it.

Suggestion: is optional on both, and it decides what you get:

  • With a Suggestion: — the entry is direct. The reviewer’s wording sits beside the original and applies in one click, like an edit.
  • Without one — the entry is advisory. The panel shows Condense this paragraph or Develop this beat in place of the prose, with the reviewer’s Why: as the direction. You rewrite the passage yourself and mark it rewritten — see Advisory suggestions.

Advisory is the honest answer when a beat needs your voice rather than the reviewer’s. The formatting instructions say so explicitly, so a reviewer working from them should offer guidance instead of inventing prose it isn’t confident in — a fabricated Suggestion: is worse than a direction, because it arrives one click from your manuscript looking finished.

Every operation entry targets one scene via SceneId:. Items in the same block may target different scenes — the importer routes each entry to its own scene.

  • IDs must be real values from the manuscript or the scene-ID list the template includes. Unmatched IDs are surfaced in the destination preview; a wrong ID that happens to name a real scene can misroute feedback, so never guess.
  • A note without an ID — every note of a book kept outside Radial Timeline, for example — is named with Note: instead, exactly as the formatting instructions list it: Note: Chapter 9. Note: works on every operation, MEMO and QUERY, so a comment from an editorial letter about one chapter lands on that chapter alone.
  • If a reviewer can’t identify the scene for a passage, the right move is to omit the SceneId and Note entirely. Editorialist then looks for the quoted text across the book and places the entry in the one scene that contains it; otherwise it routes the entry to the scene you’re currently viewing and flags it for manual verification, which is recoverable. A confidently wrong ID is not.
  • If a Radial Timeline manuscript export was the reviewer’s input, scene IDs appear inline in that export and match the template’s list. See Radial Timeline Integration.

Original: and Target: text is matched conservatively against the live note: exact text first, then a fallback that tolerates differences in quotes, dashes, spacing, and italics or bold — so a passage quoted from a Word file, where _me_ in your note arrives as plain me, is still found. Editorialist also checks whether the suggested replacement already appears to be applied. Each suggestion gets a match type — exact, multiple matches, not found, or already applied — surfaced in the review UI so you know what you’re acting on.

When the passage in your note has italics or bold and the reviewer’s text has none, applying keeps your formatting in the wording the suggestion leaves unchanged, and the card says so. If the suggestion rewrites the formatted words themselves, or would cut or move only part of a formatted phrase, the card shows the passage but does not apply it: make that change by hand, then mark it rewritten. A reviewer who writes their own italics or bold in the new text gets exactly that.


For structural work — scene-range directives, manuscript-wide design intent, a checklist the author walks through across multiple sessions — the reviewer outputs a complete markdown file instead:

---
type: editorialism
title: <Short, descriptive title>
book: <Active book name — must match the book label exactly>
reviewer: <Who the notes are from>
reviewer_type: <developmental-editor | editor | line-editor | copy-editor | agent | beta-reader | ai-editor>
source: <Optional [[wiki link]] to the letter this was distilled from>
status: in-progress
created: 2026-06-10
---
# <Same as title>
## <Theme or pillar — one section per major concern>
- [ ] Specific actionable directive [scope:: <scope>] [tags:: <tag1>, <tag2>]
- [ ] Another directive in the same theme [scope:: <scope>]
## <Another section>
- [ ] Single-scene directive [scope:: 22]
- [ ] Scene-range directive [scope:: 13–22]
- [ ] Manuscript-wide design directive [scope:: manuscript]
- [ ] Guidance the reviewer tied to no place in the book [scope:: general]
- [ ] Subplot-level work [scope:: subplot:Shail IT subplot]

Paste the reply into the review launcher: when it contains an editorialism file, the launcher shows a Import editorialism action that writes it to Editorialist/<Book>/<Title>.md (creating the folder), then opens the Editorialisms Panel. Re-importing identical content leaves the file untouched. Changed content with the same title and reviewer requires an update confirmation and retains progress on unchanged instructions; a different reviewer’s agenda is kept separately. You can still create the file by hand if you prefer — the panel picks up any type: editorialism file under Editorialist/.

Required:

  • Frontmatter type: editorialism — files without this are ignored.
  • book: must match the active book label exactly.

Attribution (recommended):

  • reviewer: — whose directives these are. The panel shows the name beside the title, and saving through the launcher adds them to the contributor directory (identity only, no stats).
  • reviewer_type: — the same role vocabulary as a review batch’s ReviewerType:.
  • source: — optional wiki link to an existing vault note holding the letter or document. The importer does not save that original document for you.

Inline metadata per item:

  • [scope:: <value>] (recommended): manuscript (work that touches every scene), general (guidance the reviewer tied to no place in the book), a scene number (22), a range (13–22, en-dash or hyphen), or subplot:<name>. A scene’s number is the number at the start of its note name (22 The Ball), or after Chapter in a book of chapter notes (Chapter 22). Anchors use the same number. The formatting instructions tell an AI to use general when the reviewer names no place, rather than turn a hint such as “earlier” into a range, and to keep manuscript for work across the whole book.
  • [tags:: tag1, tag2] (optional).
  • [words:: <n>] or [scenes:: <n>] for new prose, or [effort:: light|medium|heavy] for relative non-drafting effort (optional). These inform suggested effort, not fixed calendar commitments.
  • [decision:: <answer>] — the author’s settled answer to a directive that asks for a choice, written by Decide in the Editorialisms Panel. A reviewer never writes one; when revising or re-exporting an agenda that already carries one, keep it verbatim and treat the answer as settled.
  • [question:: <text>] — the author’s own question about a directive, written by Ask a question…; the item is marked [?]. See Questions.

Status markers (the character inside the task brackets):

Marker Status
[ ] open
[/] in progress
[x] done
[-] deferred
[?] question

The review launcher: Your editor's Word file with Choose Word file first, then Or work with an AI with two numbered steps, Send your scenes out and Bring the notes back

Run Open review launcher (command palette). The launcher modal:

  1. Checks your clipboard — a recognized batch imports directly when placement is clear; entries needing attention open a preview.
  2. Paste AI reply — if the clipboard is empty or holds something else, choose Paste AI reply under 2 Bring the notes back (in Or work with an AI) and paste; validation runs as you paste, with specific error messages. Back sets the paste aside.
  3. Destination preview — inspect placement warnings and unplaced entries. Review mis-targeted entries opens the correction step when needed. Otherwise use Import placed entries when some entries are omitted, or Import and start review when everything is placed. The action is disabled when nothing resolves.
  4. Copy instructions for AI — step 1 copies the full format guidance, both templates, and your book’s actual scene-ID list to the clipboard.

Import appends review blocks to the targeted scene notes. Nothing else in the note is touched, and no suggestion is applied until you act on it in Line edits.

Replies to the Editorialisms panel’s Draft fixes with AI… and Hand off to AI prompts are ordinary review batches and come in the same way — see Handing directives to AI.

A copy editor or line editor who works in Word sends back a .docx with tracked changes and comments. Editorialist reads that file directly, so no AI sits between your editor’s changes and your manuscript. For the whole round trip in order, from the file you send to the file you send back, see Working with a Copy Editor.

  1. Open the review launcher. Its first card is Your editor’s Word file: choose Choose Word file… and pick the .docx.
  2. The launcher says how many edits, cuts and comments it found, and whose they are. If more than one person tracked changes (your own replies in an earlier round, for example), choose whose changes to bring in under Changes by. Set Their role; it starts as Copy editor.
  3. Choose Import and start review, or Preview destinations to check where each change lands first.

Each tracked change becomes a suggestion covering the sentence it sits in, as the sentence read before the edit and as your editor left it. Changes in the same sentence come in as one suggestion. A deleted paragraph becomes a cut, and a paragraph your editor added joins the end of the paragraph above it. A comment on a changed sentence becomes that suggestion’s reason; any other comment becomes a memo on its scene. A comment above the first scene heading, such as your editor’s overall notes on a title page, becomes a note on the whole pass: it shows with every scene, and the response lists it once under General comments. Your editor is credited as the contributor, and the import joins their editorial delivery like any other.

When you apply a suggestion, only the words your editor changed are written. Your own quotation marks, dashes, italics and bold stay as they are in the rest of the sentence, even though Word shows them differently.

Italics and bold your editor added or took off come in too. A word they put in italics arrives as *word* in the suggestion, and applying it writes the italics into your scene. So do italic or bold words they typed in a new passage.

Send a file that keeps scene IDs. Export the manuscript with Radial Timeline’s Editor round option, which puts each scene’s ID in its heading. Editorialist then knows which scene every change belongs to. Without scene IDs, each change is placed by finding its sentence in your book, and a comment that is not on a changed sentence cannot be placed at all.

The launcher tells you what it leaves out, most important first. Your editor’s words that could not come in, a change it could not read or a comment that cannot be tied to a scene, sit in a warning headed Read these in Word. Layout that stays in Word folds under What stays in Word:

  • Formatting changes other than italics and bold, such as a font or size. Make these by hand.
  • Changes to headings.
  • Paragraphs split or joined. The wording comes in; the paragraph break does not.

At the end of a copy edit, your editor expects an answer: a reply to their comments and queries, and a note on any change you kept as written (a stet). Editorialist writes that response for you as a Word file.

  • Reply to a comment. Under each comment from your editor in Line edits, choose Reply.
  • Say why you kept a change. When you reject one of your editor’s changes, or mark it as rewritten, choose Say why you kept it on its card. A reason is optional.
  • Save the file. When a person’s pass is finished, the completion card leads with Write your response to (your editor). You can also run Write your response to the editor from the command palette at any point.

The response window lists every comment and every change you kept or rewrote, scene by scene. Each is a card that reads like a review thread: your passage, then your editor’s words under their name, then a box for your reply. A comment that came in on several scenes is listed once, under General comments. Replies are saved as you type, so you can close the window and come back. Save Word file writes the response to the Editorialist/Responses folder in your vault. Then Open file opens it in Word so you can read it over, and Show in Finder (Show in File Explorer on Windows) opens the folder with the file selected, ready to attach to an email. On a phone or tablet those two buttons aren’t there; send the file from your computer instead. Counts at the top show how many changes you accepted, kept, rewrote or haven’t decided yet, and how many comments you replied to, with the unanswered and undecided ones marked. The save buttons sit in a bar at the bottom of the window, so saving never needs a scroll. Each kept change shows your editor’s wording with their deleted words struck through and added words underlined, then your reason.

When your editor’s changes came in from their Word file, Editorialist keeps a copy of that file in the Editorialist/Word files folder, and the response window offers Save in their file. That writes your answers into a copy of your editor’s own document, in Editorialist/Responses, the way you would answer them in Word:

  • Your reply to a comment appears under that comment, as a reply in the same thread.
  • A change you kept as written gets a comment saying Stet, with your reason. One you rewrote says Rewritten another way. If your editor had explained the change in a comment, this appears as your reply to it.
  • Your editor’s tracked changes are left exactly as they were. Most publishers ask authors not to accept or reject a copy editor’s changes themselves; a comment saying stet is how you mark one you don’t want.

Response to Morgan Lee window: count chips, a change kept as written with the author's reason, a comment with the author's reply, an unanswered comment, and the bar at the bottom with the author's name, Done, Save as a letter and Save in their file

The first time, add your name under Name on your comments, in the bar beside the buttons. Word shows it on each of your comments. Save as a letter still makes the separate response described above, if your editor prefers that.

Write the response before you clean the review blocks out of your scenes. It is built from those blocks, so once they are cleaned it can no longer be made. Passes from an AI get no response.

The canonical formatting instructions include both formats, actual-reviewer attribution, supported effort metadata, and guidance for stable checklist updates. Copy them afresh from the launcher when starting a new review.

A memo is commentary without an individual completion checkbox. Put independently actionable developmental tasks in Editorialism checklist items; keep supporting context in memos. One handoff may contain both without duplicating the same task.

Each import joins its reviewer’s open delivery in the active book, or starts one received that day, titled from the reviewer’s role (“Developmental review”). A two-part paste lands both parts in the same delivery. A delivery stays open until its return deadline passes; one with no deadline stays open for 45 days after it was received. Imports with no reviewer, and an AI’s own reviews, join no delivery. Open … → Editorial deliveries to add the return deadline, fix a link, or record the source note, then choose Plan this delivery. See Revision Plan. Neither format creates scheduled sessions or Pending edits. Dates, priorities, prerequisites, scheduling phases, and minute ranges belong in the planner; inventing extra import fields will not set them.

When replacing an existing agenda, preserve unchanged item text, section headings, scope, and completion markers. These identify planned work. Use a distinct title for a new editorial round; deactivate older files instead of deleting them when they no longer apply.

A checklist-only paste is a valid import; no batch is required. Choose Import editorialism. Identical agendas open the existing library without rewriting the file, including when your completion markers have changed since the original import. Changed agendas show an update confirmation with added and removed instructions. Progress on uniquely matched, unchanged instructions and anchors is retained, and so is a decision or question you recorded on such an instruction when the incoming line does not state its own; changed wording or scope is treated as different work. With both formats in one paste, the launcher says the paste has two parts and gives each its own card: Matching scenes for the revision notes and Editorialism for the checklist. Importing one keeps the launcher open on the other. Reopen the launcher with the same paste and the parts already in are marked ✓ Imported and ✓ In your library, so only the missing part needs a click. The formatting instructions ask the AI to put both in one outer code block, so that block’s copy button takes both; copying the whole message works too.

Headings set in Kalice, Claude Type, designed by Margot Lévêque.