Short Drama Development

Turn the creator's intentions into a story system that continuously produces choices, costs, and state changes. Preserve the creator's choices of genre, ending, and intensity; don't treat popular formulas as story answers.

Quick Start

Offline verify precise indexing, slicing, and source file change detection for a multi-episode complete script:

bash

python3 {技能目录}/scripts/selftest.py
python3 {技能目录}/scripts/episode_intake.py index <多集整稿> --out <episode-intake-index.json>

Before You Start

This skill can be installed and run independently. First read the materials explicitly provided by the user and the direct inputs of this task; if the current directory is a

short-drama

project and project tooling is available, you may read

status

and use its release lifecycle, but missing core or any other skill is not a blocker for development work. Stage contract defines the boundaries, production-form inputs, and rule tables for this stage; no need to read other skill files.

Determine the Entry Point First

Only an idea, genre, or emotional goal: Complete direction exploration, then establish the creative brief and story engine.
Already have a synopsis, novel, adaptation materials, or series notes: Save the sources and first list committed facts, changeable ranges, source excerpts, and pending confirmations. For long materials, first perform semantic segmentation, a functional ledger, and candidate episodes; do not directly release episodes or assets based on regex name/noun matching; don't substitute the original intent.
Direction is set, only episode planning is needed: Read the existing brief and engine, then directly create or revise the episode map.
Already have a single-episode script: Don't fabricate development materials for process completeness. Writing/revising the script goes to
```
$short-drama-write
```
; asset breakdown or subsequent production can enter directly from the corresponding skills.
Already have a multi-episode complete script, need to generate or complete the episode map: Keep the original files, and only load Multi-episode Complete Script Intake and Resumable Processing. Let the Agent decide boundaries and the current batch based on the file's actual structure; tools only perform precise indexing, per-episode slicing, validation, and resumable continuation, and don't make creative judgments.

Work directly when there is enough information. Ask only when a gap would change the protagonist, conflict engine, ending promise, or adaptation boundary.

Every Run

1. Lock the Creator Contract

Read the creator's inputs and accepted files; distinguish:

Non-changeable facts, themes, characters, and ending promises;
Explorable blanks;
Form constraints and production boundaries;
Recurring emotional, informational, or power payoffs for the target audience.

For new projects, copy creative-brief.md. If changing already accepted facts, first show the semantic differences and downstream impact; don't overwrite directly.

2. Explore Genuinely Different Directions

When direction is undecided, propose a small number of candidates that differ mechanically: change the protagonist's strategy, source of antagonism, costs, or the way conflict persists, not just titles, professions, or wording. When direction is clear, deepen directly; don't manufacture fake differences to pad options.

For each candidate, explain: dramatic promise, protagonist's pursuit, antagonism mechanism, recurring payoff, upgradeable states, long-term termination condition, and major production burden. Let the creator choose or combine, and record the reasons for trade-offs; unselected options don't enter established facts.

When a complete method is needed, read story-craft.md. When the genre is already decided, you may fetch one genre card from genre-cards.md by index to calibrate the more common sources of pressure, scene granularity, and production burden for that genre; it is a reference that can be overridden with one sentence from the creator, does not replace direction selection, and doesn't enter deliverables.

If the setup has the protagonist carrying an already-occurred experience, or gaining an externally granted ability, that is a premise device layered on top of the genre; read premise-devices.md separately — a device removes friction, so it must first be given boundaries and costs. Devices don't replace genre cards, and don't get their own genre.

3. Establish the Story Engine

Copy story-engine.md, and write down clearly:

How pressure arises and how the protagonist usually responds;
How the opponent or system counterattacks, and why the protagonist can't easily walk away;
What each round of conflict changes, rather than how to repeat louder;
Each character's goals, leverage, bottom lines, and relational pressure;
What character change or revealed fact would end this engine;
Visual, sound, scene scale, and content boundaries.

For serialized dramas, long-form adaptations, or heavy vertical character change, also read serial-character-and-memory.md: first build a backstory reserve for the author's decisions only, choose a narrative entry window where old strategies start to cost, then record the character's defended beliefs, stress tests, information permissions, and cross-episode memory. Don't dump the backstory reserve directly into the first episode's dialogue.

4. Translate Series Movement into Episodes

Read episode-design.md, then copy episode-map.jsonl. For each episode, record entry state, opening, episode objective, resistance, directional turn, episode result, information release, outgoing pressure, and facts the next episode must inherit.

First ensure each episode produces a local dramatic result and adjacent episodes can hand off precisely; only then discuss the dual-track rhythm of external pressure / emotional load. Episode count, hook types, reversals, and climax placement are decided by the creator and the project.

5. Do Owner Checks and Hand Over

Before submitting a preview, check:

Referenced episodes, setups, and payoff records are findable, and unresolved items are clearly marked;
Candidates actually change mechanics, and the chosen direction runs through the brief, engine, and episode map;
Every escalation can point to a change in power, information, relationship, exposure, resources, time, or cost;
Each episode first delivers part of its promise, then leaves a specific decision, danger, or question;
No format preferences are disguised as universal rules.

These are owner self-checks, not final review. Show the creator a readable summary of additions/edits, and ask the creator to accept it before using it as a downstream source; when a quality conclusion is needed, make

$short-drama-review

a separate follow-up request, preferably executed by a reviewer who did not participate in creating the current version.

If the project needs a director's statement, this skill only produces a

project-development/director-brief.md

candidate and marks the semantic differences it intends to write into

visual_direction

production_profile

. After the creator accepts,

$short-drama

routing elevates the choice to creator authority; the development skill doesn't directly modify that authority config.

Rule Levels

structural_invariant
: Locally provable references, IDs, and explicit state contradictions; can be blocked by validators.
reviewed_invariant
: Semantic obligations such as whether an escalation is real or a promise is kept; judged by a reviewer citing evidence.
craft_default
: Practices that are usually helpful; can be overridden when the creator provides reasons.
taste_option
: Choices such as hooks, arcs, perspective, and ending temperament; cannot be used alone to block.

Don't replace causal judgment with fixed plot points, turn timings, length ratios, or quantity formulas.

Artifacts and Boundaries

This skill only owns:

```
project-development/creative-brief.md
```
```
project-development/story-engine.md
```
```
project-development/director-brief.md
```
(when the project needs it; only a creator authority candidate)
```
project-development/adaptation-map.jsonl
```
(for long material adaptation; keeps only input locator/span, de-referenced functional summaries, candidate destinations, and unresolved items; doesn't copy original text; see adaptation-map.example.jsonl)
```
project-development/series-arc.json
```
(when the project needs it)
```
project-development/episode-intake-index.json
```
(for multi-episode complete script intake; a rebuildable mechanical index, not a creative fact)
```
project-development/episode-map.jsonl
```

It doesn't write scene actions or dialogue, doesn't break down assets, doesn't write image/video prompts, doesn't generate media, and doesn't issue final review conclusions. Script semantics are handled by

$short-drama-write

Load on Demand

Promises, engine, character pressure, escalation, and setup payoff: story-craft.md
Episode contracts, causal beats, episode handoffs, and map revision: episode-design.md
Existing multi-episode complete/partial scripts; want to avoid the whole script entering context and support interruptible resumption: multi-episode-intake.md
Character drives, backstory and entry point, cross-episode changes, information permissions, dual-track rhythm, and memory recovery: serial-character-and-memory.md
Have raw material; need to compress/merge characters/scenes or visualize information: adaptation-craft.md
Creator provides reference works, sample scripts, or prompts, and wants to learn mechanics rather than imitate expression: creative-reference-intake.md
The same mechanism is repeatedly run in episodic stories, or the middle starts repeating and thinning out: mechanism-loop.md
Need to distinguish reveals, reversals, payoffs, and hooks, or select pressure mechanisms by genre: reveal-reversal-payoff.md
Genre-based qualitative case methods for choosing conflict progression, designing openings, or planning episode-end hooks: genre-and-hook-playbook.md
Genre already determined; need the more common pressure, scene granularity, hook orientation, and production difficulties for that genre: genre-cards.md (index and recall spec; read only one card at a time)
Protagonist starts with a known result, or gains an externally granted ability: premise-devices.md (device layer, layered on top of the genre card)
Draft project-level visual direction and production rules (director's statement): director-brief-craft.md
What this stage owns, inherits, and doesn't overstep: stage-contract.md

short-drama-develop

NPX Install

Tags

SKILL.md Content (Chinese)