AI 短劇編劇 structures micro-drama writing but struggles with its packaging
This prompt-engineering agent skill brings structural screenwriting discipline, explicit pacing constraints, and strict formatting rules to LLM-generated vertical dramas. However, its delivery as a localized desktop Codex skill introduces technical friction that isolates it from the very creatives it aims to serve.
Autonomously generated. This product was selected by the automated daily curation process. The jury evaluation, scores, article text, and publication were generated automatically. No human edited the jury scores or verdict before first publication.
Selection and product details
Jury Summary
The vertical-screen micro-drama industry runs on raw speed and formulaic pacing. AI 短劇編劇 solves a genuine operational pain point: standard LLMs are notorious for failing at screenwriting constraints, often outputting generic stories with vague action lines or overstepping into video-model instructions. By organizing the writing process into an eight-stage workflow covering premise, character relationships, and granular scene structure, this skill produces shootable scripts with clear subtext and cliffhangers in Traditional Chinese. Yet, the jury remains divided over its delivery vehicle. Packaging these rules as a standalone skill requiring Git, manual directory mapping, and the specialized desktop Codex CLI makes an excellent prompting framework unnecessarily difficult to access. While developers will appreciate the structured YAML and testing baselines, the actual screenwriters who need these prompts face a steep technical barrier for what is ultimately a set of sophisticated system prompts. To capture real value, this prompting logic needs to be refactored for more mainstream execution frameworks.
WHERE THE JURY AGREED
- ✓
The structural rules are highly effective at preventing LLM hallucinations, specifically stopping the AI from generating redundant video prompts or breaking formatting standards.
- ✓
The localized Traditional Chinese framing is well-suited for its target regional micro-drama markets.
- ✓
The inclusion of clear checklists and multi-stage testing validation shows an unusual level of rigor for a prompt-engineering repository.
WHERE THE JURY SPLIT
- usability onboarding
Lisa and Alex argue that requiring Git and folder copying to specialized local directories creates an insurmountable barrier for non-technical writers. Marcus and Sarah contend that within the constraints of the Codex skill ecosystem, the setup is standard and represents a deliberate architectural choice for local execution.
- differentiation insight
Marcus believes that tying the project to the obscure, local-first Codex skill ecosystem represents a strategic dead-end with limited adoption potential. David argues that treating prompt engineering with this level of structural modularity and regression testing is a novel approach that elevates prompt design into a software engineering discipline.
Five Jury Perspectives
Five simulated professional perspectives scored the same public evidence using the JuryPress Open Product Rubric.
Alex
Serial Entrepreneur
This is an effective set of guardrails for a real, booming industry that needs speed, but it is trapped in the wrong box. The actual people writing these ninety-second dramas on Monday mornings do not use terminal environments or run Git clones. If we cannot get this into a simple copy-paste format or a web tool, it is an ingenious solution that will sit on GitHub unused by its target market.
- Targets the lucrative and high-velocity vertical-screen drama market with exact format matches.
- Stops the AI from wasting generation tokens on useless video-generation prompts when only text is needed.
The local directory installation path is a massive source of friction for creative writers.
View full scorecard
Solves a real operational headache for fast micro-drama scripting by keeping the AI focused on drama elements rather than video instructions.
The prompt structures are laid out in the repository, but actual runtime execution was not verified since there is no live demo.
Not assessed by the jury: no source-code evidence was collected, so technical quality cannot be assessed from the material.
- no source-code evidence was collected, so technical quality cannot be assessed from the material
Cloning git repositories and navigating to hidden directories will immediately turn away creative writers.
Focusing on vertical drama constraints is a strategic niche that general-purpose prompts ignore.
License is fine, but there is no roadmap for adding new platform structures.
David
Principal Software Engineer
Underneath the prompt-engineering layer, the directory layout is clean and the testing methodology is remarkably structured. By using markdown files like format.md and checklists.md as reference modules, it maintains clear architectural boundaries. However, because it contains no executable code, we are entirely dependent on the host environment parsing safety and reliability.
- Outstanding directory modularity, separating core workflows, format rules, and evaluation checklists.
- Comprehensive testing documentation including baseline, behavior, and trigger microtests.
Completely lacks any automated test execution script, relying entirely on documented manual test logs.
View full scorecard
The scope addresses the structural issues of AI scriptwriting directly, framing them as technical constraints.
The YAML configuration files and directories exist, but the lack of executable runtime code or a running web app means we can only verify the file footprint.
Not assessed by the jury: no source-code evidence was collected, so technical quality cannot be assessed from the material.
- no source-code evidence was collected, so technical quality cannot be assessed from the material
Manual terminal operations are required to copy prompts, which is highly inefficient for deployment.
Converting creative screenwriting steps into strict engineering constraints is clever, but the Codex skill vehicle is highly limiting.
The project has an MIT license and is structured cleanly, but has a stagnant v0.1.0 release history.
Lisa
Head of Product Design
The user experience here is deeply conflicted. While the generated output script format is beautifully clean and respects necessary production limits, getting there requires navigating manual PowerShell or macOS copy commands. Forcing non-technical creatives to manage a hidden local directory path under the user profile is a major usability failure.
- The generated script format is perfectly tailored for vertical production, separating actions and dialogues clearly.
- Clean, explicit error-prevention rules that keep the AI on track without over-complicating the prompt.
The setup guide expects users to manually debug directory placement if the skill does not load immediately.
View full scorecard
Excellent understanding of how formatting rules prevent cognitive load on readers and actors.
I can see the documentation files, but a visual editor or playground is missing.
Not assessed by the jury: no source-code evidence was collected, so technical quality cannot be assessed from the material.
- no source-code evidence was collected, so technical quality cannot be assessed from the material
Forcing the user to manually copy directories to absolute system paths is a terrible onboarding path.
Structuring the screenwriting steps into sub-components like scene headers and dialogue tags is smart UX design for prompting.
Good license and structure, but lack of automated setup tools shows a lack of empathy for users.
Sarah
Senior Product Manager
The scope definition here is a model of product discipline. By explicitly declining to generate camera angles, video prompts, or algorithmic advice, the tool protects the creative integrity of the script and avoids prompt bloat. It does one thing, Traditional Chinese micro-drama screenwriting, and does it exceptionally well.
- A highly focused scope that explicitly rejects non-writing tasks like image-generation prompting.
- Maintains strict regional boundary compliance, focusing exclusively on Traditional Chinese micro-dramas.
No visible project roadmap or guidelines for how community contributors can submit new drama genres.
View full scorecard
Highly coherent scope. It targets the vertical-screen format and stays inside that boundary beautifully without feature creep.
The testing documentation (RED and GREEN tests) shows that the maintainers have rigorously validated their prompt triggers.
Not assessed by the jury: no source-code evidence was collected, so technical quality cannot be assessed from the material.
- no source-code evidence was collected, so technical quality cannot be assessed from the material
The setup is standard for the Codex CLI environment, though it severely limits the addressable audience.
A solid application of agentic design to scripting, though relying on Codex CLI is a limiting architectural choice.
Strict adherence to v0.1.0 release tagging and standard MIT licensing, though long-term roadmap tracking is absent.
Marcus
Venture Capitalist
The micro-drama market is growing exponentially, making this a strategic domain to target. However, by packaging this as a localized ChatGPT desktop Codex skill, the project has built a tiny island on an experimental OpenAI standard that is largely ignored by the wider developer community. To capture real value, this prompting logic needs to be refactored for more mainstream execution frameworks.
- Excellent alignment with a high-growth creative market that desperately needs content automation.
- The prompt architecture supports multi-agent workflows, facilitating future integration with video generation tools.
The reliance on the obscure Codex desktop environment limits the project's ecosystem footprint.
View full scorecard
Addresses a massive growth market in short-form video content with clear commercial utility.
The static files are all in place, but we have no telemetry on actual usage or developer API calls.
Not assessed by the jury: no source-code evidence was collected, so technical quality cannot be assessed from the material.
- no source-code evidence was collected, so technical quality cannot be assessed from the material
The niche local install method will severely drag down initial adoption curves.
While the vertical drama prompt architecture is smart, betting on the obscure Codex desktop agent ecosystem is a strategic mistake.
The repository is clean and has decent stargazers and forks (488 stars and 69 forks), but lacks a structured community governance model.
Final Verdict
Creative teams already working within terminal-based or IDE-integrated OpenAI workflows should try AI 短劇編劇 immediately to streamline their vertical drama production lines. Teams without a technical background should bypass the local installation scripts and directly extract the markdown workflows from the repository for manual use in standard web interfaces. For the project to achieve broader success, the maintainers must decouple these prompting rules from local file system dependencies. Providing a web-based playground or a direct custom GPT configuration would immediately bridge the gap between technical rigor and creative accessibility.
Evidence reach: the jury examined 0 source files, none bearing on security- or reliability-sensitive implementation paths. Not examined: execution & permission safety, data write safety, cost & resource controls, production reliability.
Bring the jury to your own project
Run the same five AI personas with your own evidence and evaluation criteria using Judgie-AI.
Explore Judgie-AI →Sources, evidence map and generation metadata
Sources
- ev-4887ef22: POUND0423/AI-drama-pound GitHub API Metadata (api_metadata)Retrieved: 2026-08-29T03:01:31.531Z
- ev-da65ad06: POUND0423/AI-drama-pound README (readme)Retrieved: 2026-08-29T03:01:31.680Z
- ev-eeeb177a: POUND0423/AI-drama-pound (official_site)Retrieved: 2026-08-29T03:01:32.694Z
What the jury could not assess
- The jury could not assess the actual script generation quality across different OpenAI model versions, as execution runs were not part of the examined repository.
- The security implications of local agents execution could not be verified due to the lack of raw codebase execution in the evidence reach.
How claims relate to sources
After this review was written, a separate pass recorded how its statements relate to the collected material. It is a record of the writing, not a score of it: opinions and comparisons are expected to be the jury's own.
This record covers the review's narrative — the summary, headline, standfirst, jury summary, points of agreement and disagreement, stated limitations, verdict, and each judge's verdict and leading concern — plus any specific factual claim made elsewhere, such as a figure, a security or runtime assertion, or a claim about what the project lacks. The per-criterion scoring commentary is not mapped statement by statement: an opinion about a score is the jury's judgment, not a claim about the world. All 65 covered statements were recorded.
- Repository observation10 statements
- Creator claim11 statements
- Editorial judgment44 statements
Statements recorded as more than one claim
These sentences assert more than one thing, and the collected material does not cover every part equally. Each part is recorded separately so that a well-sourced half does not stand in for the whole. Where the parts differ, the statement is counted at the strength of its weakest factual part.
- “An agent-based prompting skill structured for the ChatGPT desktop Codex environment, designed to output tightly constrained vertical-screen drama scripts in Traditional Chinese.”
- An agent-based prompting skill structured for the ChatGPT desktop Codex environment
- designed to output tightly constrained vertical-screen drama scripts in Traditional Chinese.
- “It restricts generations to production-ready formats, preventing common AI scripting issues like bloated scene descriptions and lack of dialogue subtext.”
- It restricts generations to production-ready formats
- preventing common AI scripting issues like bloated scene descriptions and lack of dialogue subtext.
- “Packaging these rules as a standalone skill requiring Git, manual directory mapping, and the specialized desktop Codex CLI makes an excellent prompting framework unnecessarily difficult to access.”
- Packaging these rules as a standalone skill requiring Git, manual directory mapping, and the specialized desktop Codex CLI
- makes an excellent prompting framework unnecessarily difficult to access.
- “While the generated output script format is beautifully clean and respects necessary production limits, getting there requires navigating manual PowerShell or macOS copy commands.”
- While the generated output script format is beautifully clean and respects necessary production limits
- getting there requires navigating manual PowerShell or macOS copy commands.
Generation metadata
- Model: gemini-3.5-flash
- Prompt version: 4.7.0
- Rubric: open-source-product 2.0.0
- Scores recalculated by code: yes
- Editorial provenance: Autonomously generated
- Evidence record: complete — 65/65 covered statements (32 scoring statements out of scope)
Discuss this review
Disagree with the verdict or found evidence we missed? Share a reasoned response, public evidence, or a factual correction.
Comments are public and require a GitHub account. Comments do not automatically change the jury score. Verified corrections may be reflected separately in Corrections & Updates.
Open GitHub Discussions