# Horror Craft

Horror is the manipulation of fear through structure, not just imagery.
The three modes of fear must be ordered deliberately.

## Dread vs. terror vs. gross-out

- **Dread** — the anticipation of horror. The reader knows (or suspects)
  something is wrong before the character does. Dread is the most powerful
  and most sustainable mode; it carries whole novels.
- **Terror** — the moment of confrontation. Short, sharp, and expensive:
  terror spent too often becomes noise.
- **Gross-out** — visceral disgust. The cheapest mode and the fastest to
  fatigue. Use sparingly, and never as a substitute for dread.

**Ordering rule:** build on dread, spend terror rarely, and earn gross-out
only after dread has done the work. A horror story that opens with
gross-out has nowhere to escalate.

## The uncanny

The uncanny — the familiar made wrong — is horror's highest-leverage tool:

- Take something ordinary (a nursery rhyme, a family dinner, a commute) and
  introduce one wrong detail. The reader's own familiarity does the
  frightening.
- Plan uncanny beats in the scene outline: name the familiar thing and the
  single wrongness. More than one wrongness per beat becomes surrealism,
  which is a different effect.

## Monster rules, stated early

Whatever the horror is — creature, ghost, curse, human evil — its **rules
must be established before they are exploited**:

- What it wants, what it can and cannot do, what protects against it (if
  anything). The reader must be able to feel the characters making
  *informed* bad decisions, not arbitrary ones.
- Rules may be *discovered* by the characters (that's plot), but they must
  be *consistent*: a monster that breaks its own rules for a scare has
  spent the reader's trust.
- Record the rules in the relevant worldbuilding system file and link them
  from the arc — rule changes are continuity changes (`story continuity`
  territory).

## Recovery periods

Fear needs contrast. After a terror beat, give the reader a **recovery
period**: a quieter scene where characters (and readers) process, regroup,
or experience false safety. Recovery scenes are where sequels (see
`scene-craft/references/scene-sequel.md`) do their best work in horror —
the dilemma hits harder when the reader just caught their breath.

- False safety is a tool, not a cheat: the recovery must be *genuinely*
  safer, so its violation lands. A recovery that was obviously a trap
  retroactively cheapens the quiet scene.
- Never stack two terror beats without recovery between them.

## Horror audit (revision)

- [ ] Dread carries the middle; terror is spent rarely and deliberately.
- [ ] The horror's rules are stated before they are exploited, and never broken.
- [ ] At least one uncanny beat per act (familiar thing + single wrongness).
- [ ] Every terror beat is followed by a recovery period.
- [ ] The ending's fear resolves the dread the opening planted — the thing
      the reader feared in chapter one is the thing confronted at the end.

# MG and YA Craft

Middle Grade (roughly ages 8–12) and Young Adult (roughly 12–18) are defined
less by subject matter than by **protagonist age, voice, stakes, and the
role of adults**. These are checklist-friendly constraints — apply them at
story-init and audit in revision.

## Word-count norms

Know the target before drafting; length is part of the contract with the
category's readers and gatekeepers:

- **MG:** roughly 30,000–55,000 words (fantasy may run to ~70,000).
- **YA:** roughly 55,000–90,000 words (fantasy/SF to ~100,000).

Record the target in `story.md` (`target-words:`) and check progress with
`story wordcount .`. A 130,000-word "MG" is a miscategorized adult novel —
recategorize or cut.

## Age-appropriate voice and stakes

- **Protagonist age** anchors the category: MG protagonists are typically
  10–13; YA protagonists 15–18 (rarely younger than 14). The protagonist
  should be at the *upper* end of the target readership's age — readers read
  up.
- **Voice** must read as the protagonist's age: concerns, humor, and
  emotional vocabulary calibrated to that age, not an adult looking back.
  (Voice calibration itself belongs to the `better-writing` skill; this is
  the structural constraint: *whose* age the voice must match.)
- **Stakes** are age-sized but *felt* at full intensity: losing a best
  friend, being seen as a fraud at school, a parent's disappointment. The
  fate of kingdoms works in MG/YA only when routed through what it means
  to *this* young protagonist. Never condescend to the stakes — to the
  character, they are everything.

## Minimized adult involvement

The young protagonist must be the agent of the story:

- Adults may advise, obstruct, or be absent — but they must not **solve**
  the central problem. An adult who swoops in to fix the climax breaks the
  category contract.
- Absent, incompetent, or constrained adults are structural tools: boarding
  schools, summer camps, and neglectful guardians exist to clear the field.
  If adults are present and competent, explain in-story why *they* can't act
  (and make that explanation airtight — young readers audit this).
- **Audit question:** "Could a competent adult in this world solve the
  protagonist's problem in one scene?" If yes, restructure.

## Content boundaries

- Match the category's norms for violence, romance, language, and thematic
  darkness — not as censorship but as *reader-trust*: the book was handed to
  a 10-year-old (or a 15-year-old) by someone who trusted the category.
- Darkness is allowed; hopelessness is not. MG especially requires the
  ending to affirm agency — the world may be hard, but the protagonist's
  actions mattered.

## MG/YA audit (revision)

- [ ] Protagonist age sits at the upper end of the target readership.
- [ ] Word count within category norms (or recategorized deliberately).
- [ ] No adult solves the central problem; adult absence is explained.
- [ ] Stakes are age-sized and treated with full seriousness.
- [ ] The ending affirms the protagonist's agency.

# Mystery: Fair Play and the Clue Ledger

The fair-play doctrine is the genre's contract with the reader: the reader
must have every clue the detective has, in time to solve the mystery
themselves. A mystery that withholds essentials is not clever — it is
cheating.

## Fair-play rules

- **The reader gets every clue.** Every fact the detective uses to solve the
  case must appear on the page before the solution, early enough to matter.
- **The culprit is introduced early** — conventionally within the first
  third. A culprit who appears only at the reveal is a stranger, not a
  suspect.
- **No withheld essentials.** The detective may not know something the
  reader doesn't (no secret evidence revealed at the gather-suspects scene).
  The detective may *misinterpret* — that is the game — but the facts must
  be shared.
- **The solution must be the best explanation of the planted clues**, not
  merely a consistent one. If three solutions fit the clues, the planting
  was too thin.

## Clue-planting techniques

- **Hidden in plain sight:** the clue appears as set dressing or a throwaway
  detail. The reader sees it; only the detective's context makes it a clue.
- **Camouflaged:** the clue is embedded in a longer list or a busy scene —
  one wrong alibi among five correct ones, one inconsistent detail in a
  full witness statement.
- **Misinterpreted:** the clue is correctly observed but wrongly explained,
  by the detective or by a witness the reader trusts. The reveal reframes it.
- **Absence as clue:** the dog that didn't bark — something that *should*
  be present and isn't. Plant the expectation early so the absence registers.

## Red-herring discipline

- Every red herring must be **plausible** (a reasonable reader could suspect
  this person), **scene-fitting** (it belongs in the scenes it appears in —
  never inserted solely to mislead), and **resolved, not abandoned**.
- A resolved red herring gets its own mini-payoff: the suspect is cleared
  *for a reason that teaches the reader something true* about the case.
- Track herrings in the clue ledger with `type: red-herring` and a
  `cleared-in` chapter. An uncleared herring is a broken promise.

## The gather-suspects reveal

The classic reveal structure: assemble the suspects, walk through the
evidence, eliminate each in turn, name the culprit. Requirements:

- Every elimination must use **planted** clues (check the ledger).
- The culprit's motive, means, and opportunity must each trace to a planted
  clue — three separate ledger entries minimum.
- The scene works best when the detective is *wrong about one thing* on the
  way to being right — it keeps the reveal from feeling like a lecture.

## The clue ledger

Maintain `continuity/clues/` with one file per clue (kebab-case ids). The
CLI creates them with `story add clue`:

```shell
story add clue "The silver locket" --planted chapter-02 --payoff chapter-05
```

Omit `--payoff` when the payoff is not yet known. The generated frontmatter:

```yaml
---
title: The silver locket
status: planned        # planned | planted | paid-off
planted: chapter-02
payoff: chapter-05
significance-delayed: true   # reader sees it before understanding it
characters: [edwin-marsh, priya-okafor]
arcs: [the-drowned-witness]
---
```

Add a `type: clue | red-herring` field by hand when the entry is a herring,
plus `cleared-in: chapter-NN` recording where it gets resolved with a reason.

- `planted` must precede `payoff` — `story continuity` reuses the
  promise-ordering machinery for the clue ledger: payoff before plant is an
  error, a clue planted ≥3 chapters ago with no payoff is a warning.
- A clue with `significance-delayed: true` is fair play *only if* the clue
  itself was visible; delayed significance is the game, hidden clues are
  the cheat.
- After adding or editing clues, run `story reindex .`, `story links .`,
  `story validate .`, `story continuity .`

## Mystery audit (revision)

- [ ] Every clue the solution uses is planted before the reveal.
- [ ] The culprit appears in the first third.
- [ ] Every red herring is cleared with a reason.
- [ ] No essential fact first appears in the reveal scene.
- [ ] The detective's wrong turns use misinterpretation, not missing facts.

# Romance: Beats and the HEA Contract

IP note: romance beat structure is widely discussed in craft communities
(Gwen Hayes' *Romancing the Beat* is the best-known codification). This is
a paraphrase of widely-published beat concepts, not a reproduction of any
commercial beat list.

## The reader contract: HEA / HFN

Romance makes the genre's hardest promise: a **Happy Ever After** or **Happy
For Now** ending for the central couple. This is non-negotiable — a romance
that ends with the couple apart or ambiguously together has broken its
contract, whatever else it achieved. Every structural choice serves this
promise: the reader must believe, at the end, that these two people have
earned each other.

- The HEA/HFN must be **earned by the characters' growth**, not by
  circumstance removing the obstacle. If the only thing keeping them apart
  was a misunderstanding, the romance is thin.
- The ending proves the *theme* of the relationship: what these two people
  learned about love that they couldn't learn alone.

## Beat concepts (paraphrased)

The widely-published romance arc, in structural terms:

1. **Setup / stuck in the lie** — each lead is shown living the false
   belief that keeps them from love (see the `theme-craft` lie/truth
   machinery: romance is often a dual change-positive arc).
2. **The meeting** — the leads collide under pressure. First impressions
   are wrong in an interesting way; attraction and resistance coexist.
3. **No-way / first turning point** — a reason the relationship cannot
   work is established. This is the engine: every scene tests it.
4. **Deepening** — forced proximity or shared stakes; the leads see each
   other's competence, humor, or wound. Intimacy grows *despite* the no-way.
5. **Midpoint shift** — the relationship changes state: a first kiss, a
   confession, a betrayal, a choice. After this, the old dynamic is gone.
6. **Retreat / darkening** — the lie reasserts itself; one or both leads
   pull back. The cost of vulnerability becomes concrete.
7. **The black moment** — all seems lost for the relationship: the lie wins
   a decisive round, the leads separate, the no-way looks permanent. This
   is the romance's all-is-lost beat and it must feel *final*.
8. **The turn** — the lead(s) confront the lie, usually at cost, and choose
   the truth. The grand gesture belongs here — and it must be *specific to
   this couple*, not a generic romantic set piece.
9. **HEA/HFN** — the couple together, the lie dead, the relationship's
   theme proven in a final image.

## Craft rules

- **The obstacle must be internal as well as external.** Circumstance
  (distance, a rival, a misunderstanding) can complicate, but the *real*
  barrier is what each lead believes about love — otherwise the black
  moment resolves by logistics, not growth.
- **Both leads change** (in the standard dual-arc romance). A romance where
  only one lead grows is a different story wearing romance's clothes.
- **The black moment must threaten the relationship, not just the plot.**
  Danger to life and limb raises stakes; only danger to the *bond* earns
  the beat.
- **Intimacy scenes (at any heat level) advance character**, not just the
  relationship status. Each one should change what the leads know about
  each other.

## Romance audit (revision)

- [ ] The HEA/HFN is present and unambiguous.
- [ ] The black moment feels final on first read.
- [ ] The ending is earned by growth, not by circumstance.
- [ ] Each lead has a lie, a wound, and a truth (see `theme-craft`).
- [ ] The "no-way" reason is tested in scenes, not just stated.

# The Sci-Fi Pipeline: Worldbuilding → Plot

In science fiction, the speculative element (the technology, the alien
biology, the altered physics, the future history) must be **load-bearing**:
remove it, and the plot collapses. A story that would work unchanged as a
contemporary drama with the serial numbers filed off is not science fiction
— it is a drama wearing a costume.

## The load-bearing test

For the central speculative element, ask: *if I replaced this with its
mundane equivalent, would the plot still function?*

- A faster-than-light drive that only changes the scenery fails the test.
  A faster-than-light drive whose travel time creates the story's central
  dilemma (messages outlive senders; colonies diverge) passes it.
- Apply the test at the **premise** level (story-init) and again at the
  **audit** level (revision): list three plot turns that are *impossible*
  without the speculative element. Fewer than three means the element is
  decorative — either wire it into the plot or cut it.

## Rules stated before exploited

The speculative element's rules are plot machinery, so they follow the same
discipline as horror's monster rules:

- **State the rules before the climax exploits them.** The reader must be
  able to anticipate the solution's shape, even if not its specifics. A
  rule revealed at the moment of use is a deus ex machina with better
  production values.
- **Costs and limits are more interesting than capabilities.** A technology
  that does everything generates no plot; a technology with a price, a
  range limit, or a side effect generates all of it. Record costs and
  limits in the system file, not just powers.
- **Rule changes are continuity changes.** If the draft discovers a better
  rule mid-book, update the worldbuilding system file and re-check every
  earlier scene that assumed the old rule (`story continuity` territory).

## The pipeline

Wire the speculative element through the existing machinery in this order:

1. **System file** (`worldbuilding/systems/`): rules, costs, limits,
   failure modes, who understands it.
2. **Character file**: who depends on it, who fears it, whose lie it feeds
   (see `theme-craft`) — the element should intersect at least one
   character's arc.
3. **Arc file**: at least one plot turn that is impossible without it
   (the load-bearing test, made concrete).
4. **Timeline**: the element's introduction precedes its exploitation —
   check `planted-in` vs `paid-off-in` ordering as with any setup.

## Common failure modes

- **The lecture chapter:** a character explains the science to someone who
  should already know it. Deliver rules through use, failure, or conflict
  (see `scene-craft/references/exposition.md`), not seminars.
- **The magic button:** a capability introduced exactly when needed.
  Audit: every capability used in the climax must appear (even in passing)
  in the first half.
- **The unexamined implication:** a world-changing technology with no
  social consequences (immortality nobody comments on, teleportation with
  traffic jams). If the element is load-bearing for plot, its second-order
  effects should appear in the worldbuilding — or the world feels thin.

## Sci-fi audit (revision)

- [ ] The load-bearing test: three plot turns impossible without the element.
- [ ] Rules stated before exploited; costs and limits recorded.
- [ ] No capability first appears in the climax.
- [ ] The element intersects at least one character's arc.
- [ ] Second-order effects visible in the worldbuilding.

# Serial / Episodic Structure

Serial fiction (web serials, episodic novels, seasonal releases) needs
structure at **two layers**: the within-installment layer this reference
covers, and the book/season layer the `series-continuity` skill handles
(one project per book, canon carried across). Do not confuse them.

## The season/volume goal

Every season (or volume, or series-arc) needs one **overarching goal**: the
thing the whole season is about. It should be statable in one sentence and
visible from the first installment:

- "Take the city back from the Hollow Men."
- "Find out who is killing the oath-bound."

The season goal gives installments their direction and the finale its
payoff. Record it in `story.md` (`season-goal:`) or the top of
`plot/_index.md`. A serial without a season goal drifts — readers can feel
the absence by installment five.

## The per-episode dramatic question

Every installment must pose and **answer** one dramatic question — not the
season question, a smaller one:

- Installment 3: "Will Mara get the ledger out of the counting house?"
- Installment 7: "Who warned the Hollow Men?"

Rules:

- The question is posed early in the installment and answered by its end.
  An installment that only advances the season goal without resolving its
  own question feels like a fragment.
- The answer may raise a *new* question (chains are good), but it must not
  merely defer the original one.
- Record each installment's question in the chapter frontmatter
  (`episode-question:`) and its answer in the chapter notes.

## A reward in every installment

Serial readers pay per installment (in money or attention). Each one must
deliver something complete:

- A revelation, a reversal, a set piece, a relationship shift, a mystery
  answered — at least one **concrete payoff** per installment, even while
  larger threads stay open.
- The reward should be *proportional* to the installment's position: early
  installments reward with momentum and intrigue; late installments reward
  with payoff.
- Audit: list each installment's reward in one line. An installment with
  no reward is a candidate for merging with its neighbor.

## Recap discipline

Serial installments release apart in time; readers forget. But recaps that
read as recaps insult the reader:

- **Weave, don't dump:** re-establish essential facts through the
  installment's action (a character checks the wound from last time; a
  rival references the betrayal).
- **One-paragraph maximum** of explicit "previously" material, and only
  for facts the installment's plot *requires*. Everything else is the
  reader's problem — trust them.
- New-reader onboarding is a feature, not a bug: each installment should
  orient a new reader within its first page (who, where, what's at stake)
  without punishing returning readers. See
  `scene-craft/references/openings.md`.

## Arc-specific stakes (preventing endless escalation)

Serials die by escalation inflation: every installment bigger than the last
until the story is absurd. Instead, **vary the stake type** per arc:

- Personal stakes (a relationship), professional stakes (a position),
  physical stakes (safety), moral stakes (a principle) — rotate them.
- A quiet installment about a betrayal can land harder than a loud one
  about an explosion, *because* of the contrast. Plan the stake rotation
  in the season outline.

## Serial audit (per season)

- [ ] One-sentence season goal, visible from installment one.
- [ ] Every installment poses and answers its own dramatic question.
- [ ] Every installment delivers at least one concrete reward.
- [ ] Recaps are woven, not dumped; new readers can orient in one page.
- [ ] Stakes rotate in type; no three consecutive installments escalate
      the same stake.
- [ ] The season finale pays off the season goal *and* the longest-running
      installment questions.

# Thriller Craft

Thrillers run on pressure: a protagonist outmatched, a clock running, and
chapters that end before the reader can put the book down.

## The ticking clock (honored once promised)

- If the story promises a deadline — a bomb, a trial date, a ransom window —
  the deadline is **sacred**. The climax must engage the clock: beat it,
  miss it with consequences, or reveal the clock was misread (which must
  itself be a planted, fair twist — see `mystery-fair-play.md`).
- A promised clock that fizzles ("actually there was plenty of time") is a
  broken promise. Log every deadline in `continuity/promises/` with its
  promised chapter and its payoff chapter.
- **Clock discipline:** track story-time explicitly in `plot/timeline.md`.
  The reader does thriller math — if the bomb has "twenty minutes" and the
  next three chapters take an hour of story time, the contract breaks.

## Power imbalance

The thriller protagonist starts **outmatched**: fewer resources, less
information, worse position than the opposition. The imbalance is the
engine — every scene is the protagonist spending ingenuity against superior
force.

- Record the imbalance in the arc file: what the opposition has that the
  protagonist doesn't.
- The protagonist's wins must come from **asymmetric moves** (cleverness,
  sacrifice, exploiting the opposition's blind spot), never from suddenly
  matching the opposition's strength.
- Escalate by *narrowing options*, not just raising stakes: the thriller
  tightens (fewer allies, fewer exits, less time) rather than merely
  getting louder.

## Set pieces

Plan 2–4 **set pieces**: extended, high-craft sequences (a chase, an
infiltration, a siege, a con) that the reader will remember. Each set piece:

- Has its own try/fail chain (see `scene-craft/references/try-fail.md`)
  with at least three beats.
- Changes the board: after a set piece, the situation is materially
  different (an ally lost, a resource spent, knowledge gained).
- Is *earned* by setup: the location, the tools, and the opposition's
  capabilities were all planted earlier.

## Mini-cliffhanger chapter endings

End thriller chapters on **unresolved motion**: a door opening, a phone
ringing with bad news, a decision forced but not yet made. Not every
chapter — relentless cliffhangers flatten — but the default thriller
chapter ends mid-tension, not on resolution. See
`chapter-writing/references/writing-guidelines.md` on chapter endings;
thrillers bias hard toward the cliffhanger end of that spectrum.

## Structure pairing

Thrillers pair naturally with the **Fichtean curve** pacing shape: the story
opens at or near the first crisis with minimal setup, then escalates
through a rising series of crises. (Fichtean curve is not one of the five
structure models in `plot-structure/references/structure-models.md` — when
using it, note the choice in `plot/_index.md`'s structure section and track
crises as the primary beat unit.) Cross-link the Fichtean crisis beats to
the ticking-clock deadlines in the timeline.

## Thriller audit (revision)

- [ ] Every promised deadline is engaged by the climax.
- [ ] Story-time math holds across the clock chapters.
- [ ] The protagonist never wins by suddenly matching the opposition.
- [ ] Each set piece changes the board.
- [ ] Chapter endings default to unresolved motion.