scriveno 2.0.5

package/templates/CONTEXT.md

@@ -0,0 +1,56 @@
+---
+creative_pillar: continuity
+always_load_for: [next, resume-work, pause-work, save]
+authority: bootstrap
+---
+
+# Project context
+
+This file is auto-regenerated by Scriveno. It is the single one-page bootstrap any new session (human or AI) reads first to locate itself in the project. Do not hand-edit -- changes are overwritten on the next save / pause / resume.
+
+If this file is missing, stale, or contradicts STATE.md, run `/scr:resume-work` (or `/scr:scan --fix`) to regenerate it.
+
+**Updated:** {{LAST_UPDATED}}
+**Updated by:** {{LAST_COMMAND}}
+
+---
+
+## You are here
+
+**Project:** {{TITLE}}
+**Work type:** {{WORK_TYPE}}
+**Phase:** {{PHASE}}
+**Active unit:** {{CURRENT_UNIT}}
+
+## Recent activity
+
+The last five recorded actions, newest first. The full chronological log lives in `.manuscript/HISTORY.log`; STATE.md holds the structured snapshot.
+
+| When | Command | Unit | Outcome |
+|------|---------|------|---------|
+{{RECENT_ACTIONS}}
+
+## Open items
+
+**Drafted but not reviewed:** {{UNITS_PENDING_REVIEW}}
+**Planned but not drafted:** {{UNITS_PENDING_DRAFT}}
+**Voice-check warnings since last drafter run:** {{VOICE_WARNINGS}}
+**Continuity flags:** {{CONTINUITY_FLAGS}}
+**Unresolved scaffold markers:** {{SCAFFOLD_MARKERS}}
+
+## Record highlights
+
+{{RECORD_HIGHLIGHTS}}
+
+## Suggested next step
+
+{{NEXT_STEP}}
+
+Run `/scr:next` to act on the suggestion, or `/scr:scan` first if you suspect any of the recorded state above is out of date.
+
+## Trust check
+
+This file was generated from STATE.md, OUTLINE.md, RECORD.md when present, the contents of `.manuscript/drafts/body/`, and the last 5 lines of HISTORY.log. The numbers above are claims, not facts -- a `/scr:scan` confirms they still match the disk.
+
+**Last `/scr:scan` run:** {{LAST_SCAN}}
+**Last drift verdict:** {{LAST_SCAN_VERDICT}}

package/templates/OUTLINE.md

@@ -0,0 +1,36 @@
+---
+creative_pillar: structure
+always_load_for: [discuss, plan, next]
+authority: project
+---
+
+# Outline
+
+## High-level structure
+
+{{HIGH_LEVEL}}
+
+## Act/part breakdown
+
+{{ACTS}}
+
+## {{UNIT_LABEL}} list
+
+{{UNITS}}
+
+## Arc positions
+
+**Opening:** {{OPENING_ARC}}
+**Inciting incident:** {{INCITING}}
+**First turning point:** {{TURN_1}}
+**Midpoint:** {{MIDPOINT}}
+**Second turning point:** {{TURN_2}}
+**Climax:** {{CLIMAX}}
+**Resolution:** {{RESOLUTION}}
+
+## Notes
+
+{{NOTES}}
+
+---
+*Update this file as the structure evolves. It's the bridge between WORK.md (high concept) and individual unit planning.*

package/templates/RECORD.md

@@ -0,0 +1,68 @@
+---
+creative_pillar: record
+always_load_for: [discuss, plan, draft, editor-review, next, scan, continuity-check]
+authority: established
+---
+
+# Record
+
+This file tracks what the work has established so far. It is not the manuscript, and it is not workflow state.
+
+- `STATE.md` tracks where the project is in the workflow.
+- `OUTLINE.md` tracks how the work is structured.
+- `RECORD.md` tracks what the work has established and what future units must honor.
+
+Use this file across work types. In fiction it can hold story facts. In nonfiction it can hold argument movement. In sacred commentary it can hold interpretive claims and source-handling obligations. In technical writing it can hold system facts, procedure truths, and reader expectations.
+
+## Established on page
+
+Facts, claims, decisions, events, procedures, definitions, relationships, images, or constraints that have already appeared in drafted text.
+
+| Established item | First established | Current status | Notes |
+|------------------|-------------------|----------------|-------|
+| | | | |
+
+## Open threads
+
+Questions, tensions, mysteries, arguments, promises, risks, procedures, motifs, images, relationship shifts, or reader expectations that remain open.
+
+| Thread | Opened in | Needs payoff or handling | Urgency |
+|--------|-----------|--------------------------|---------|
+| | | | |
+
+## Promises and payoffs
+
+Reader-facing promises the work has made and whether they have paid off.
+
+| Promise | Setup location | Payoff location | Status |
+|---------|----------------|-----------------|--------|
+| | | | |
+
+## Continuity facts
+
+Durable facts future drafting must not contradict.
+
+| Fact | Applies to | Source location | Notes |
+|------|------------|-----------------|-------|
+| | | | |
+
+## Movement
+
+Character, figure, subject, argument, reader, doctrine, procedure, image, object, setting, or relationship movement that future units should preserve.
+
+| Moving element | Starting state | Current state | Next pressure |
+|----------------|----------------|---------------|---------------|
+| | | | |
+
+## Next unit obligations
+
+Specific obligations the next planned or drafted unit should honor.
+
+-
+
+## Notes
+
+Use plain language. Keep this file short enough to load. Move detailed material into the relevant source file when it becomes permanent: CHARACTERS.md, THEMES.md, WORLD.md, QUESTIONS.md, PROCEDURES.md, DOCTRINES.md, or another adapted context file.
+
+---
+*This file is the living record of what the work has established. Update it after drafting, review, import, and major continuity decisions.*

package/templates/STATE.md

@@ -0,0 +1,50 @@
+---
+creative_pillar: continuity
+always_load_for: [next, scan, save, pause-work, resume-work]
+authority: state
+---
+
+# Workflow state
+
+**Last updated:** {{TIMESTAMP}}
+**Current phase:** {{PHASE}}
+**Current unit:** {{CURRENT_UNIT}}
+
+## Progress
+
+- **Units total:** {{UNITS_TOTAL}}
+- **Units discussed:** {{UNITS_DISCUSSED}}
+- **Units planned:** {{UNITS_PLANNED}}
+- **Units drafted:** {{UNITS_DRAFTED}}
+- **Units reviewed:** {{UNITS_REVIEWED}}
+- **Units submitted:** {{UNITS_SUBMITTED}}
+- **Total words:** {{WORD_COUNT}}
+
+## Last actions
+
+| Timestamp | Command | Unit | Outcome |
+|-----------|---------|------|---------|
+| | | | |
+
+## Session metrics
+
+**Current session started:** {{SESSION_START}}
+**Units this session:** {{SESSION_UNITS}}
+**Words this session:** {{SESSION_WORDS}}
+**Quality passes:** {{SESSION_QUALITY_PASSES}}
+
+## Pending
+
+- **Next step:** {{NEXT_STEP}}
+- **Open revisions:** {{OPEN_REVISIONS}}
+- **Unresolved notes:** {{UNRESOLVED_NOTES}}
+- **Voice-check issues:** {{VOICE_ISSUES}}
+- **Continuity flags:** {{CONTINUITY_FLAGS}}
+
+## Session handoff
+
+**Last session ended:** {{LAST_SESSION_END}}
+**Resume context:** {{RESUME_CONTEXT}}
+
+---
+*This file is the source of truth for workflow position. Every Scriveno command reads it and updates it. Don't edit by hand -- it's managed automatically.*

package/templates/STYLE-GUIDE.md

@@ -0,0 +1,121 @@
+---
+creative_pillar: voice
+always_load_for: [draft, voice-check, plan, editor-review, translate]
+authority: sovereign
+---
+
+# Style guide -- Voice DNA
+
+*This file is loaded into every drafter agent invocation. It is the single most important file in the project -- the difference between prose that sounds like the writer and prose that sounds like generic AI.*
+
+## Part 1 -- Narrative architecture
+
+### Narrative perspective
+- **POV:** {{POV}} (first person, close third, distant third, omniscient, second person)
+- **POV consistency:** {{POV_CONSISTENCY}} (single POV, multi-POV by chapter, shifting)
+- **Narrator reliability:** {{RELIABILITY}} (reliable, unreliable, ambiguous)
+- **Narrator distance:** {{DISTANCE}} (intimate, observational, detached, cinematic)
+
+### Tense
+- **Primary tense:** {{TENSE}} (past, present, mixed)
+- **Tense shifts:** {{TENSE_RULES}} (where and why tense shifts are allowed)
+
+### Narrative stance
+- **Knowing vs. discovering:** {{STANCE}} (does the narrator know what's coming, or discover it alongside the reader?)
+- **Emotional distance:** {{EMO_DISTANCE}} (warm/cool, sympathetic/critical/neutral)
+- **Judgment:** {{JUDGMENT}} (moralistic, withholding, ironic, compassionate)
+
+## Part 2 -- Sentence and paragraph architecture
+
+### Sentence architecture
+- **Average sentence length:** {{SENTENCE_LENGTH}} words
+- **Sentence variation:** {{VARIATION}} (high variation, mostly short, mostly long, rhythmic alternation)
+- **Complex structures:** {{COMPLEX}} (embedded clauses, parallel structures, fragments)
+- **Sentence music:** {{MUSIC}} (staccato, flowing, punchy, lyrical)
+
+### Paragraph architecture
+- **Paragraph length:** {{PARAGRAPH_LENGTH}} (short, medium, long, variable)
+- **White space:** {{WHITESPACE}} (dense, generous, rhythmic)
+- **Paragraph breaks:** {{BREAK_RULES}} (when to break -- beat shift, POV shift, emphasis)
+
+## Part 3 -- Vocabulary and figurative language
+
+### Vocabulary
+- **Register:** {{REGISTER}} (formal, conversational, lyrical, colloquial, literary)
+- **Complexity:** {{COMPLEXITY}} (accessible, elevated, specialized)
+- **Word origin:** {{WORD_ORIGIN}} (preference for Anglo-Saxon vs. Latinate)
+- **Jargon handling:** {{JARGON}} (how domain-specific terms are introduced)
+- **Profanity:** {{PROFANITY}} (none, mild, moderate, explicit, character-specific)
+
+### Figurative language
+- **Metaphor density:** {{METAPHOR_DENSITY}} (sparse, moderate, dense)
+- **Metaphor style:** {{METAPHOR_STYLE}} (grounded/concrete, abstract/philosophical, sensory)
+- **Recurring image systems:** {{IMAGE_SYSTEMS}} (list of motifs that should recur)
+- **Similes:** {{SIMILES}} (frequency and style)
+- **Symbolism:** {{SYMBOLISM}} (overt, buried, emergent)
+
+## Part 4 -- Dialogue
+
+### Dialogue voice
+- **Ratio:** {{DIALOGUE_RATIO}} (% of text that is dialogue)
+- **Tag style:** {{TAG_STYLE}} (said-bookish, creative tags, action beats, minimal)
+- **Subtext level:** {{SUBTEXT}} (on-the-nose, moderately layered, heavily subtextual)
+- **Dialect/accent:** {{DIALECT}} (phonetic, light, none)
+- **Interrupted speech:** {{INTERRUPTIONS}} (frequency and handling)
+
+### Character voice differentiation
+Each character should sound distinct. See CHARACTERS.md for individual voice anchors.
+
+## Part 5 -- Description and sensory detail
+
+### Description density
+- **Overall density:** {{DESCRIPTION_DENSITY}} (sparse, moderate, rich, lush)
+- **Sense mix:** {{SENSES}} (visual-dominant, multi-sensory, synesthetic)
+- **Specificity:** {{SPECIFICITY}} (general, specific, hyper-specific with brand names)
+
+### Physicality
+- **Body awareness:** {{PHYSICALITY}} (how often characters' physical sensations appear)
+- **Place:** {{PLACE}} (how strongly setting is felt)
+- **Time of day/weather:** {{ATMOSPHERE}} (how often atmosphere is referenced)
+
+## Part 6 -- Pacing, rhythm, and transitions
+
+### Pacing
+- **Default pace:** {{PACE}} (slow-burn, steady, brisk, breakneck)
+- **Pace variation:** {{PACE_VARIATION}} (where and how pace shifts)
+- **Scene-to-summary ratio:** {{SCENE_SUMMARY}}
+
+### Transitions
+- **Between scenes:** {{SCENE_TRANSITIONS}} (hard cuts, soft fades, hooks)
+- **Between chapters:** {{CHAPTER_TRANSITIONS}} (cliffhangers, echoes, clean breaks)
+- **Time jumps:** {{TIME_JUMPS}} (signaling, frequency, handling)
+
+## Part 7 -- Reference influences
+
+### Authors/works the writer wants to evoke
+{{INFLUENCES}}
+
+### Passages loaded as reference
+{{REFERENCE_PASSAGES}}
+*(500-word samples from the writer's own previous work or selected reference authors, used by the drafter as voice anchors.)*
+
+## Part 8 -- Sacred voice registers (sacred/historical work types only)
+
+{{SACRED_REGISTERS_BLOCK}}
+
+*If the work type is sacred/historical, this section lists the active voice registers (prophetic, wisdom, legal, liturgical, narrative-historical, apocalyptic, epistolary, psalmic, parabolic, didactic) and which units use which register.*
+
+## Part 9 -- Do and don't
+
+### Always
+{{ALWAYS}}
+
+### Never
+{{NEVER}}
+
+### Consider
+{{CONSIDER}}
+
+---
+
+*This style guide is generated from the `/scr:profile-writer` questionnaire and refined by `/scr:voice-test`. It is updated whenever the writer says "this doesn't sound like me" or flags a specific voice issue. Every drafter agent loads this file -- it is the single most important artifact in the project.*

package/templates/THEMES.md

@@ -0,0 +1,36 @@
+---
+creative_pillar: themes
+always_load_for: [discuss, plan, draft, editor-review]
+authority: project
+---
+
+# Themes
+
+*The thematic threads this work explores. Each theme has a position (the writer's stance), a set of scenes where it surfaces, and a craft strategy for handling it.*
+
+---
+
+## {{THEME_NAME}}
+
+**The question:** {{QUESTION}}
+**The writer's position:** {{POSITION}} (or "deliberately unresolved")
+**How it surfaces:** {{HOW}} (dialogue, action, imagery, structure, contrast)
+**Scenes where it appears:** {{SCENE_LIST}}
+**Craft strategy:** {{STRATEGY}} (show without telling, contrast through character, embedded in setting, etc.)
+**Risks:** {{RISKS}} (preachiness, obviousness, loss of subtlety)
+
+---
+
+## Subject dynamics
+
+Use this section when the work does not depend on named characters, or when a unit is driven more by an idea, subject, procedure, place, object, doctrine, argument, reader problem, or image pattern than by cast dynamics.
+
+### {{SUBJECT_NAME}}
+
+**Subject type:** {{SUBJECT_TYPE}} (idea, claim, process, procedure, place, object, doctrine, image pattern, reader problem)
+**Reader starts here:** {{READER_START}}
+**Reader should move toward:** {{READER_SHIFT}}
+**Pressure or friction:** {{SUBJECT_PRESSURE}} (misconception, counterclaim, risk, constraint, ambiguity, failure mode)
+**Interaction pattern:** {{INTERACTION_PATTERN}} (claim vs. counterclaim, rule vs. exception, step vs. failure mode, image vs. meaning, doctrine vs. practice)
+**How it should surface:** {{SURFACE_STRATEGY}} (structure, example, evidence, image, rhythm, demonstration, contrast)
+**Watchpoints:** {{SUBJECT_WATCHPOINTS}}

package/templates/WORK.md

@@ -0,0 +1,67 @@
+---
+creative_pillar: work
+always_load_for: [discuss, plan, next]
+authority: project
+---
+
+# {{WORK_TITLE}}
+
+## Premise
+
+{{PREMISE}}
+
+## Work type
+
+- **Type:** {{WORK_TYPE}}
+- **Group:** {{GROUP}}
+- **Command unit:** {{COMMAND_UNIT}}
+- **Target length:** {{TARGET_LENGTH}}
+- **Genre:** {{GENRE}}
+- **Audience:** {{AUDIENCE}}
+
+## Profile
+
+<!--
+  The `tradition` key applies only to sacred work types.
+  The `platform` key applies to work types that publish as books (prose, visual, poetry, sacred).
+  For work types where neither applies (script, academic, technical, interactive, speech_song),
+  `/scr:new-work` renders this section as "Not applicable for this work type." and omits the keys.
+  Valid values and per-work-type defaults live in `data/CONSTRAINTS.json` under `architectural_profiles`.
+-->
+
+{{PROFILE_BLOCK}}
+
+## Elevator pitch
+
+{{ELEVATOR_PITCH}}
+
+## Logline
+
+{{LOGLINE}}
+
+## Comparable works
+
+{{COMPARABLES}}
+
+## Central question
+
+{{CENTRAL_QUESTION}}
+
+## Core conflict
+
+{{CORE_CONFLICT}}
+
+## Tone and mood
+
+{{TONE}}
+
+## Structural notes
+
+{{STRUCTURE_NOTES}}
+
+## What makes this different
+
+{{DIFFERENTIATORS}}
+
+---
+*This file is the north star. Every drafted unit should advance the premise, sharpen the central question, and serve the tone. If you find yourself off-track, return here.*

package/templates/WORLD.md

@@ -0,0 +1,62 @@
+---
+creative_pillar: world
+always_load_for: [discuss, plan, draft, continuity-check]
+authority: project
+---
+
+# World
+
+*The setting rules that constrain everything. If it's not in here, don't assume it exists.*
+
+---
+
+## Geography
+
+**Scale:** {{SCALE}} (city, region, continent, planet, multiverse)
+**Key locations:**
+{{LOCATIONS}}
+
+**Maps/spatial notes:**
+{{SPATIAL_NOTES}}
+
+---
+
+## Culture
+
+**Societies:** {{SOCIETIES}}
+**Social structure:** {{SOCIAL_STRUCTURE}}
+**Languages:** {{LANGUAGES}}
+**Customs and traditions:** {{CUSTOMS}}
+**Religion/belief systems:** {{BELIEFS}}
+
+---
+
+## Technology / Magic
+
+**Tech level:** {{TECH_LEVEL}}
+**Key technologies:** {{KEY_TECH}}
+**Magic system (if applicable):** {{MAGIC_SYSTEM}}
+**Rules/limitations:** {{RULES}}
+**Cost of magic/tech:** {{COST}}
+
+---
+
+## Rules and Laws
+
+**Governing systems:** {{GOVERNMENT}}
+**Laws that matter to the story:** {{LAWS}}
+**Enforcement:** {{ENFORCEMENT}}
+**What's forbidden:** {{FORBIDDEN}}
+**What's common but surprising to outsiders:** {{SURPRISING}}
+
+---
+
+## History
+
+**Timeline of events relevant to the story:** {{TIMELINE}}
+**Key historical events:** {{EVENTS}}
+**How the past shapes the present:** {{PAST_TO_PRESENT}}
+**Historical conflicts still active:** {{ACTIVE_CONFLICTS}}
+
+---
+*Update this file as the world develops. Build progressively -- seed questions first, then refine with `/scr:build-world --area <area>`.*

package/templates/WRITING-RULES.md

@@ -0,0 +1,156 @@
+# Writing rules
+
+*Universal pitfalls every drafter must avoid. Loaded into the drafter, voice-checker, and originality-check after STYLE-GUIDE.md. STYLE-GUIDE.md wins where they conflict; this file is the floor, not the ceiling.*
+
+*This file is optional. If absent, agents fall back to their built-in rules. If present, this is the single source of truth for universal AI-tell don'ts.*
+
+## How to read this file
+
+- **Defaults, not absolutes.** If STYLE-GUIDE.md says the writer hedges, fragments dialogue, or moralizes deliberately, follow STYLE-GUIDE.md.
+- **Floor, not ceiling.** These rules catch the worst AI tells. Real craft goes further.
+- **Per work type.** When a per-work-type pitfall pack exists for the project's `work_type`, load it after this file. Type-specific rules can refine but not relax universal ones.
+
+## Universal don'ts
+
+These patterns make prose sound generated. Avoid them unless STYLE-GUIDE.md explicitly calls for them.
+
+### Human-first restraint
+- Do not over-correct prose that already sounds like the writer.
+- Isolated signals are not enough. Look for clusters: generic vocabulary plus symmetrical rhythm plus vague claims plus formulaic transitions.
+- Preserve human markers: specific details, mixed feelings, fragments, self-corrections, uneven rhythm, and era-bound references.
+- If the passage has only one or two weak spots, fix those spots. Do not sand down the whole paragraph.
+- When reporting edits, name one to three things deliberately left alone because they are authentic to the writer or register.
+
+### Variance over substitution
+- Do not fix AI-sounding prose by swapping one suspect word for another everywhere. Fix the underlying thought.
+- Increase or preserve genuine structural variance: sentence length, paragraph shape, syntax, pace, and silence.
+- Do not install a new humanizer signature. Avoid repeated edit shapes such as every paragraph becoming short punch, long explanation, short punch.
+- Match STYLE-GUIDE.md's distribution of rhythm. If the writer is deliberately even, spare, maximalist, clipped, or formal, do not force generic variety.
+
+### Factual integrity and content preservation
+- Do not invent facts, names, numbers, sources, quotes, dates, prices, examples, locations, or claims.
+- Specificity is good only when the source material provides it.
+- Cover every original beat the plan or draft requires. Do not truncate, skip paragraphs, compress away obligations, or replace concrete content with a prettier generalization.
+- If a detail is missing, leave the gap visible for the writer or planner rather than filling it with plausible material.
+- Audit soft inference as carefully as hard facts. Do not add causal, temporal, quantitative, or priority claims unless the source or plan states them.
+
+### Register-aware restraint
+- Academic, technical, legal, sacred, journalistic, and quoted material should not be casualized just to sound "human."
+- Preserve required terms of art, citation language, doctrinal phrasing, legal precision, procedural commands, and period diction.
+- Change register only when STYLE-GUIDE.md or the plan asks for the shift.
+- When a sentence is correct for its register, leave it alone even if it would sound warmer in another context.
+
+### Stance discipline
+- Add edge, opinion, irony, warmth, or first-person attitude only when STYLE-GUIDE.md, the plan, or the writer explicitly calls for that stance.
+- Stance may react to content already present. It may not smuggle in new facts, motives, causes, events, expertise, lived experience, or claims.
+- If the writer asks for more voice, make the voice sharper through rhythm, emphasis, and judgment about supplied material, not through invented support.
+
+### Hedging and qualifiers
+- Avoid "perhaps", "maybe", "in a sense", "to some degree", "it could be argued", "one could say", "it bears mentioning", "it is worth noting", "it should be noted".
+- Avoid stacked qualifiers: "quite", "rather", "somewhat", "fairly", "relatively", "arguably", "potentially".
+- A sentence that hedges twice is hedging too much.
+
+### Throat-clearing and scaffolding
+- Do not open with "The scene begins...", "In this chapter...", "What follows is...".
+- No "and then" connective tissue between beats.
+- No meta-commentary on the prose itself.
+- Start in the moment. If you cannot start, re-read the previous unit's tail and let its rhythm lead.
+
+### Balanced-both-sides constructions
+- Do not pair every pro with a con, every advantage with a disadvantage.
+- Real voices take positions. Symmetry is an AI tell.
+- If STYLE-GUIDE.md describes the voice as "judicious" or "essayistic", balance is allowed. Otherwise, lean.
+
+### Generic metaphors and dead figures
+- "Heart of gold", "tip of the iceberg", "at the end of the day", "wave of emotion", "shiver down the spine", "time stood still".
+- If the metaphor predates the writer, it is not the writer's metaphor.
+- Match metaphor density and image systems to STYLE-GUIDE.md.
+
+### Symmetrical rhythm
+- Do not write three sentences of similar length in a row.
+- Vary cadence: short, short, long. Or fragment, full, full.
+- Uniform paragraph lengths are an AI tell. Real rhythm breathes.
+
+### Moralizing closings
+- Do not wrap the unit in a bow. No takeaway sentence, no "and so", no lesson.
+- The scene ends where it ends. The reader does the work.
+- If STYLE-GUIDE.md establishes a moralizing voice (homiletic, didactic, parabolic), defer to it.
+
+### Essay transitions in narrative
+- "Furthermore", "moreover", "additionally", "in conclusion", "consequently" do not belong in fiction or scene.
+- They belong in argument. Use them only when the work type's group is `academic` or `technical`, or when the unit is explicitly an essay.
+
+### Abstract vagueness
+- "Various factors", "a number of reasons", "in many ways", "something like", "some kind of".
+- Name the factors. Name the reasons. Name the thing.
+- Specificity is voice. Vagueness is filler.
+
+### Chat artifacts and placeholder contamination
+- Remove chatbot wrapper text: "I hope this helps", "Would you like me to", "Here is the revised version", "As an AI".
+- Remove copied citation artifacts, UI references, orphaned markdown fences, access-date placeholders, and template blanks.
+- Do not leave `[INSERT]`, `[YEAR]`, `[COMPANY]`, `[TODO]`, `turn0search0`, `oai_citation`, or similar residue in writer-facing prose.
+- If the artifact hides missing information, flag the missing information instead of inventing it.
+
+### Emotional telling
+- Do not write "she felt sad", "he was angry", "they were excited".
+- Show it through action, dialogue, body, or implication.
+- The verb "felt" before an emotion word is a flag. Recheck before keeping it.
+
+### AI tics in dialogue
+- Characters do not "let out a sigh", "give a small smile", or "nod softly".
+- They sigh, smile, or nod. Or they do something more specific.
+- Adverb stacks on tags ("she said softly, quietly, almost inaudibly") are a tell. One adverb at most. Usually none.
+
+### Dialogue attribution defaults
+- Default to "said". "Said" is invisible.
+- Action beats are stronger than creative tags. "He set down the cup." beats "he muttered darkly".
+- Attribute only when the speaker would be unclear. Two-character scenes need fewer tags than you think.
+
+### Diff-anchored explanation
+- In docs, comments, and technical prose, describe what is true now, not what changed from an older version.
+- Release notes and changelogs may describe changes. Durable documentation should stand on its own.
+- Prefer "Authentication uses encrypted cookies for session storage" over "We replaced the old session middleware with encrypted cookies."
+
+## Show-don't-tell triggers
+
+Before writing any of these, try the harder version first.
+
+| Trigger phrase | What it usually means | Try instead |
+|---|---|---|
+| "She felt X" | Emotion told, not shown | A gesture, a thought, a physical sensation |
+| "He realized X" | Insight stated flat | Let the reader watch the realization land |
+| "It was a beautiful day" | Setting told | One concrete detail of weather, light, or air |
+| "The room was tense" | Mood told | Body language, pause, what is not said |
+| "She knew that X" | Cognition stated | Show what she does because she knows it |
+
+## Punctuation defaults
+
+- No em dashes. No en dashes. Use commas, colons, semicolons, parentheses, or two sentences.
+- No emojis in prose, dialogue, or headings.
+- Hyphens for compounds ("old-fashioned") and number ranges ("pages 10-15").
+
+## Diagnostic discipline (honest read)
+
+These rules tell a writer how to write. When the work is instead *diagnosed* for AI-slop (`/scr:voice-check`, `/scr:originality-check`), the same rules read from the other side, under this discipline:
+
+- **Diagnose, do not rewrite.** A diagnostic reports a band, a score, and flagged spans with reasons. It never hands back an improved or suggested version of a span. Fixing flagged prose is a separate transform step (`/scr:line-edit`, `/scr:polish`, re-draft) the writer chooses, then the diagnostic re-runs as a fresh read. The loop is diagnose, decide, transform, re-verify; keeping the halves apart prevents a score-then-rewrite gaming loop, so never carry a target score into the rewrite.
+- **Uniformity is the signal.** Prose reads as AI because of sameness (even lengths, even rhythm, the same shapes resolved the same way), not vocabulary. Flag the signature, not the lone word. A relocated signature (words swapped, rhythm still even) is not more authentic and does not earn back points.
+- **Restraint over reach.** Match scrutiny to evidence density; low density biases hard toward a high score and a near-empty flag list. A lone weak signal that does not recur or co-occur is not a flag and must not lower a score. A strong false positive that is a real human marker (concrete detail, mixed feeling, dated reference, self-correction, unhedged opinion, trade idiolect, a STYLE-GUIDE.md tic) is positive evidence and moves the score up. Over-flagging genuine human prose is the worst error a diagnostic can make.
+- **No diagnostic signature.** Do not always flag the first sentence, force a fixed flag count, or score to a safe middle. Vary the verdict with the evidence.
+- **Honest read, no detector.** A diagnostic is an honest read of how authentically the prose reads as the writer's own work. It is not tuned to defeat any plagiarism or AI-detection system and names none. Reframe "get this past a graded or contractual check" toward the honest diagnostic.
+
+## When STYLE-GUIDE.md overrides
+
+STYLE-GUIDE.md takes precedence when it explicitly establishes:
+- A hedging or qualified voice (philosophical, essayistic, scholarly registers)
+- Fragmented or symmetrical rhythm as a deliberate signature
+- A moralizing or didactic closing pattern (homiletic, parabolic, catechetical work types)
+- Genre-specific stock language the writer wants preserved (pastiche, parody, period voice)
+- Profanity, dialect, or register choices that would otherwise read as AI artifacts
+- A clear stance profile: skeptical, warm, devotional, polemical, restrained, comic, severe, or otherwise writer-specific
+
+When STYLE-GUIDE.md is silent, this file's defaults hold.
+
+---
+
+*If a sentence sounds like a smart machine wrote it, it probably did. Rewrite until it sounds like the writer.*