refactor(fiction-assistance)!: drop line-editor agent, add proofread and check-consistency commands

Two shifts in scope:

- Remove the line-editor agent. The plugin no longer mutates the
  manuscript — all commands report findings to the conversation
  and leave revisions to the writer.

- Consolidate the consistency surface. /check-voice and
  /check-continuity were narrow checks with overlapping concerns;
  they are replaced by /check-consistency, a single broader pass
  covering author voice, character voice, words-vs-actions
  alignment, and in-world believability. Plot-logic and
  sequencing concerns formerly in /check-continuity are now
  served by /generate-timeline's consistency-issues section.

- Add /proofread for the mechanical layer that previously had no
  command equivalent — grammar, tense, punctuation, spelling,
  agreement, sentence-level clarity. Reports issues; does not edit.

Command count unchanged at nine. The plugin is now agent-free and
mutation-free; every command is read-only on the manuscript.

BREAKING CHANGE: /check-voice and /check-continuity are removed
(superseded by /check-consistency). The line-editor agent is
removed (use /proofread for reporting; the writer makes the edits).

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
This commit is contained in:
movq
2026-05-16 18:38:28 -05:00
parent 6e4edc09d1
commit 187863fa16
7 changed files with 126 additions and 138 deletions

View File

@@ -0,0 +1,63 @@
---
description: Consistency and believability pass — author voice, character voice, behavior matches words, in-world coherence
argument-hint: "[path, glob, or selection]"
allowed-tools: [Read, Glob, Grep]
---
Read the material at `$ARGUMENTS` and assess whether the piece holds together: stable voices, characters who act like themselves, a world whose rules and inhabitants stay coherent. This is distinct from developmental review (which asks *"is this a good story?"*) — this asks *"is this a coherent story?"* Report findings; do not edit.
## What to assess
### 1. Author / narrator voice
Across the whole piece (and across scenes if they differ):
- **Diction and register** — does the narrator's word choice and formality stay stable? Where does it shift, and is the shift earned (POV change, time jump, emotional escalation) or accidental?
- **Sentence rhythm and shape** — does the narrator's characteristic cadence hold?
- **Attitudinal stance** — irony, warmth, distance, sympathy. Does the narrator's relationship to the material drift unintentionally?
- **Tense and POV discipline** — slips between close-third and distant-third, between past and present, between focal characters within a scene
### 2. Character voice
For each character with meaningful dialogue or interior:
- **Distinctiveness** — could a reader identify the speaker with tags removed? Test by sampling.
- **Consistency** — does the voice hold across scenes and emotional registers?
- **Overlap** — any two characters bleeding into each other?
- **Growth versus drift** — does change in the character show up in the voice deliberately, or does the voice just wander?
### 3. Words versus actions
For each character of consequence:
- **Stated values vs. choices** — does the character's behavior match what they say they believe, want, or fear? When the gap is deliberate (hypocrisy, self-deception), is it earning its space? When the gap looks accidental, flag it.
- **Stated capabilities vs. demonstrated** — does the supposedly-shrewd character get fooled too easily? Does the supposedly-physical character avoid physical solutions? Either is a flag.
- **Relationships vs. behavior** — do characters treat each other consistently with the relationship the text claims? Reported intimacy that doesn't show up in the prose, or coldness contradicted by warm gestures, are signals.
- **Decisions under pressure** — would *this character*, as established, actually make *this choice* here? Or is the plot pushing them?
### 4. In-world believability
- **Rules of the world** — physical, magical, technological, social, economic. Do the rules established earlier hold later? Where they're bent, is it earned (revelation, breakthrough, exception with a cost) or convenient?
- **Cause and consequence** — do actions have appropriate weight? Do injuries persist, do reputations propagate, do resources deplete, do secrets stay secret in proportion to who knows them?
- **Setting plausibility** — given the world, do crowds behave like crowds, institutions like institutions, frontier towns like frontier towns? Or is the setting a backdrop that responds only to the protagonist?
- **Coincidence and contrivance** — are improbable events seeded, or do they appear when the plot needs them? "And then…" beats that should be "but…" or "therefore…".
- **Information economy** — does each character act on what they reasonably know, not on what the reader knows? Does plot-required ignorance feel real, or do characters fail to notice the obvious so the story can proceed?
## Output format
Per concern, when issues are found:
- **Type** (author voice / character voice / words-vs-actions / world coherence / believability)
- **Location** (chapter/scene with a short quoted excerpt)
- **Issue** — what conflicts with what, with both references
- **Severity** (minor — attentive readers might catch / moderate — careful readers will / major — breaks immersion)
- **Why it matters** — what trust the issue costs the reader
- **Suggested fix** — a concrete option, not just "address this"
Close with:
- **What's holding together well** — name the strongest threads of consistency so the writer doesn't sand them off in revision
- **Priority fixes** — three to five concrete suggestions in order of impact
- **Patterns to watch** — recurring slips the writer should look for in their own future drafts (e.g., "voice flattens in action scenes," "secondary characters share a register")
Be honest but not pedantic. Distinguish between unearned drift (a problem) and deliberate variation (a feature). When in doubt, note the ambiguity rather than the verdict.

View File

@@ -1,41 +0,0 @@
---
description: Check a manuscript for continuity problems — plot holes, character inconsistencies, world-building contradictions
argument-hint: "[path, glob, or selection]"
allowed-tools: [Read, Glob, Grep]
---
Read the manuscript at `$ARGUMENTS` (a file, a glob of chapters, or pasted text) and identify continuity issues. Report what you find directly to the conversation.
## What to check
### Character consistency
- **Traits and personality** — does the character act consistently across scenes?
- **Physical description** — height, eye color, age, scars, clothing — any contradictions?
- **Knowledge** — do they know something in chapter 4 that they shouldn't learn until chapter 7?
- **Relationships** — do established relationships develop consistently or shift without cause?
- **Speech patterns** — does their voice stay recognizable?
- **Motivations** — do their goals stay aligned with what's been established?
### World-building consistency
- **Rules of the universe** — physics, magic, technology limits, social structures
- **Geography and locations** — distances, layouts, travel times
- **History and timeline** — past events referenced consistently
- **Cultural and political details** — group behaviors, hierarchies, allegiances
### Plot logic
- **Information flow** — how does each character learn what they know? Track it
- **Cause and effect** — does each consequence have a setup?
- **Object and detail persistence** — items, wounds, weather, time of day
- **Pacing of revelations** — anything resolved before it was a problem, or set up and never paid off
## Output format
For each issue found:
- **Type** (character / world / plot)
- **Location** (chapter/scene, with quoted context where possible)
- **Issue** (what contradicts what, with both references)
- **Severity** (minor — reader probably won't catch / moderate — attentive readers will / major — breaks the story)
- **Suggested fix** (a concrete option, not just "address this")
Close with a short summary of the most load-bearing issues — the three to five the writer should fix first.

View File

@@ -1,41 +0,0 @@
---
description: Analyze character voice — consistency, distinctiveness, and development
argument-hint: "[path, glob, or selection]"
allowed-tools: [Read, Glob, Grep]
---
Read the manuscript at `$ARGUMENTS` and analyze how distinct and consistent each character's voice is. Report findings to the conversation.
## What to analyze
For each named character with meaningful dialogue:
### Speech profile
- **Vocabulary** — formal/informal, technical/plain, characteristic word choices
- **Sentence shape** — short/long, complete/fragmented, complex/simple
- **Rhythm** — clipped, flowing, interrupted, deliberate
- **Emotional register** — how they show anger, fear, affection, humor
- **Background markers** — regional, generational, professional, cultural cues that appear in speech
### Consistency
- Does the voice hold across scenes and chapters?
- Does it adapt appropriately under stress, with different conversational partners, or across an arc — or does it slip and re-form without reason?
- Does any growth in the character show up *in the voice* rather than only in the narration?
### Distinctiveness
- Could a reader identify the speaker with the dialogue tags removed? Test by sampling unattributed lines.
- Are any two characters bleeding into each other? Name the overlap.
- Are any characters' voices so close to the narrator's that they don't stand out?
## Output format
Per character (focus on the major ones):
- **Voice profile** — 2-3 sentences capturing the signature
- **Sample lines** — 2-3 quoted lines that exemplify the voice
- **Consistency** — strong / mostly stable / drifts; cite the strongest example and any slip
- **Distinctiveness** — distinctive / shared traits with [other character] / undifferentiated
Close with:
- **Overall assessment** — which voices are working, which need work
- **Priority fixes** — three to five concrete suggestions ordered by impact

View File

@@ -0,0 +1,59 @@
---
description: Proofread and line-edit pass — grammar, tense, punctuation, spelling, mechanical errors. Reports issues; does not edit.
argument-hint: "[path, glob, or selection]"
allowed-tools: [Read, Glob, Grep]
---
Read the text at `$ARGUMENTS` and identify mechanical and sentence-level problems. Report what you find — **do not edit the file**. The writer is making the changes themselves.
## What to check
### Grammar and mechanics
- **Verb tense** — consistency within a sentence, paragraph, and across the piece; unintended slips between past and present
- **Subject-verb agreement** — number, person, indefinite pronouns
- **Pronoun-antecedent agreement** — number, gender, unclear reference
- **Modifiers** — dangling, misplaced, squinting
- **Parallelism** — in lists, comparisons, correlative conjunctions
### Punctuation and conventions
- **Commas** — splices (unless deliberate), missing commas after introductory phrases, serial comma consistency
- **Semicolons, colons, dashes** — used correctly and consistently
- **Quotation marks and dialogue tags** — punctuation inside/outside, comma vs. period before tags, paragraph breaks per speaker
- **Apostrophes** — possessive vs. plural, *its* vs. *it's*
- **Hyphenation** — compound modifiers, hanging hyphens, en/em dash usage
- **Capitalization** — proper nouns, titles, after colons, in dialogue
### Spelling and word choice
- **Misspellings** — including names that drift across the manuscript
- **Homophones** — their/there/they're, your/you're, peek/peak/pique, etc.
- **Confused words** — affect/effect, fewer/less, who/whom, lie/lay, etc.
- **Repeated words** — accidental doubles, awkward proximity repetition
### Sentence-level clarity
- **Sentence fragments** — unintentional vs. stylistic; flag the unintentional ones
- **Run-ons and comma splices** — flag unless clearly deliberate for voice
- **Awkward constructions** — passive where active would be clearer, nominalization piles, tangled syntax
- **Word order** — modifier placement, sentence rhythm
- **Tightening opportunities** — redundancies, throat-clearing phrases, weak intensifiers (very, really, quite, just), filter words (saw, heard, felt, noticed) when they distance the reader
### Style consistency
- **Spelling/style variants** — US vs. UK, *grey/gray*, *-ize/-ise*, etc. Pick one and flag drift
- **Number formatting** — when to spell out, when to use digits
- **Time, date, and unit formatting**
- **Italics, bold, small caps** — consistency for thoughts, emphasis, foreign words, titles
## Output format
For each issue:
- **Location** — chapter/scene/line if known, with a short quoted excerpt
- **Issue type** — one of the categories above
- **What's wrong** — concrete diagnosis
- **Suggested fix** — show the corrected version, not just the rule
Group issues by type so the writer can attack one category at a time. Close with:
- **Summary** — total issues by category
- **Patterns** — recurring problems the writer should watch for in future drafts (e.g., "consistent tense-slip in interior-thought passages")
Be a careful reader, not a pedant. Don't flag stylistic choices that read as deliberate — fragments for voice, comma splices for rhythm, dialogue that breaks rules to sound right. Note them only if a choice looks accidental.