short-drama-review

Original：🇨🇳 Chinese

Translated

2 scripts

Reviews source analysis, story, script, visual design, image prompts, storyboards, frozen keyframes, video prompts, and existing media in short-drama projects. Used when the user asks to "review/check script", "check assets or continuity", "check image/video prompts", "check source analysis", "review template feel", or "calibrate the project based on production observations". Only writes review issues, conclusions, and revision requirements; it does not modify source files on behalf of the owner.

15installs

Sourcezenstory-ai/drama-skills

Added on2026-09-25

NPX Install

npx skill4agent add zenstory-ai/drama-skills short-drama-review

SKILL.md Content (Chinese)

View Translation Comparison →

Short Drama Review

Prefer to have a reviewer who did not participate in the creation of the current version perform the review; when conditions do not allow, self-review is acceptable and should be disclosed honestly. Review and revision are two separate work units: this skill only identifies issues, impacts, and revision results that must be achieved; it does not modify sources on behalf of the owner in the same round.

Quick Start

Read the current file named by the user directly. When a single-episode review needs to be persisted, write

review/EP001-review.md

; for standalone file reviews, use

review/<topic>-review.md

. If the user only wants verbal conclusions, respond directly; do not create a review file for the sake of "completeness".

Suggested structure for review Markdown:

markdown

# EP001 Review

- Scope: script, storyboard, video prompts
- Conclusion: REVISE
- Review method: independent reviewer / self-check

## Blocker · REV-001 · On-screen text is not carried by any shot
- Location: script.md / EP001-SC002; storyboard.md / SHOT-EP001-004
- Evidence: ...
- Impact: ...
- Required revision: add accurate text and legibility conditions via short-drama-storyboard.
- Rule: SHT-01 · reviewed_invariant

Citations use only file names, heading IDs, line numbers, or short quotes; do not create sources declarations, hashes, record IDs, or a second set of states.

Choosing the Review Scope

```
source_analysis
```
```
story_script
```
```
assets_continuity
```
```
image_prompts
```
```
storyboard_keyframes
```
```
video_prompts
```
```
production_outputs
```
```
full_episode
```
```
delivery_privacy
```
```
project_calibration
```

Read only the corresponding materials; do not preload the entire project:

Source analysis: Source Analysis Review Rubric
Story, scenes, actions, and dialogue: Story & Script Review Rubric
Identity, variants, continuity, and image prompts: Asset Prompt Review Rubric
Source text delivery, shots, keyframes, and video motion: Visual Motion Review Rubric
Complete review methodology: Review Method
Common production-side defects: Production Quality Gates
In-project calibration from authorized production observations: Project Calibration
Template feel, repetitive techniques, or AI flavor: Anti-Template Repair
Stage boundaries, reference media, and rule tables: Stage Contract

Workflow

1. Freeze the Scope

Clearly write down which files, sections, shots, or prompts are under review, along with the constraints the creator has explicitly stated. If the target changes during the review, re-read before drawing conclusions.

2. Check Provable Facts First

Whether the visible IDs, citations, and durations across the five documents are consistent;
Whether script dialogue, actions, sounds, and on-screen text all have visual/audio carriers;
Whether characters, locations, prop states, and shot start/end points are continuous;
Whether prompts run from an accurate start point to an accurate end point;
When the target model produces synchronized audio and video, whether the audio content, timing, and vocal boundaries are consistent with the source; whether dialogue-free shots specify room tone or silence (
```
VID-22
```
,
```
VID-23
```
, see Visual Motion Review Rubric);
Whether private inputs, credentials, absolute paths, or internal process text are leaked.

When necessary inputs are missing, only pause the judgments that depend on them; other issues can still be aggregated.

3. Review Content with Evidence

Each finding includes: location, necessary short quotation or conflicting facts, audience/production impact, the revision result that must be achieved, owner, severity, and rule level. Do not merely say "AI flavor", "not cinematic enough", or give scores without evidence.

4. Cross-Document Synthesis

text

Script facts -> visual design -> shot responsibility and boundaries -> frozen keyframes -> video motion -> next state

Prioritize preserving intent, information timing, and continuity; do not reward flashy prompts detached from the source.

5. Conclusion and Assignment

```
APPROVE
```
: no blocking issues;
```
APPROVE_WITH_NOTES
```
: only non-blocking improvements;
```
REVISE
```
: structural, content, or constraint conflicts exist;
```
PROVISIONAL
```
: key inputs are insufficient; judgment cannot be completed for now.

Group revision requirements by owner. Return control after this round; revision and re-review begin only when the user explicitly requests them.

Severity

```
blocker
```
: unsafe, undeliverable, or would send the process in the wrong direction;
```
major
```
: clearly damages story comprehension, continuity, or production results;
```
minor
```
: has concrete impact but does not block;
```
note
```
: creative choice or optional polish.

Boundaries

Do not submit image, video, TTS, or music tasks, do not configure adapters, and do not treat Dashboard operations as production authorization.
When media cannot be read, state the limitation explicitly; do not infer facial consistency, performance, lip sync, mixing, or market performance from text or adapter states.
Production observations must be tied to exact inputs, prompts, references, configurations, and result versions; they must not be generalized into universal rules.
Review files retain only the minimal evidence needed for revision and do not copy complete private inputs.

Installation and Maintenance

Run

python3 scripts/selftest.py

only during installation, upgrade, or troubleshooting.