$ show cclank/lanshu-awesome-ai-video-kit
This repository is a meticulously curated knowledge base and toolkit for AI video prompt engineering. Its core is a structured data pipeline: raw knowledge from official vendor documentation and YouTube masterclasses is reverse-engineered into standardized methodologies (21+ SOPs) and a centralized prompt database (`prompts/data/all-prompts.json`). This data is then exposed through a cross-model benchmark (`prompts/data/cross-model-matrix.json`), zero-dependency web tools (Prompt Browser, Cross-Model Matrix), and executable Claude Code Skills (`skills/`). The architecture is data-centric. The primary value lies in the curated, structured data assets and the rigorous methodology used to create them. The application layer (web tools, skills) is a thin, well-crafted consumer of this data. The repository actively manages its primary riskβmodel version driftβthrough an automated monitoring system (`scripts/monitor_models.py`) that hashes external endpoints and files issues on change. While primarily a data_research project, it exhibits strong application characteristics through its interactive tools and AI skill integrations. It is a living knowledge base, actively maintained with a clear contribution workflow and a strong emphasis on data provenance and reproducibility. The `cross-model-matrix.json` is a standout asset, providing a controlled benchmark for comparing prompt styles across 11 major AI video models.
The LLM identified 4 files worth highlighting for reuse:
A meticulously structured JSON dataset containing 110 prompts (10 scenarios x 11 models). It uses a `core_anchor` pattern to keep the scene intent identical while varying the prompt structure per model's best practices. This is a highly reusable research artifact for comparing AI video models, training prompt translators, or studying model behavior. It is self-contained, well-documented, and has a clear schema.
A Python script that monitors external model documentation/API endpoints for changes using SHA-256 hashing. It is configured via `model_endpoints.yaml` and tracks known hashes in `known_hashes.json`. This pattern is directly reusable for any project that needs to track changes in external, versioned resources (e.g., API docs, model cards, pricing pages). It is cleanly separated from the rest of the codebase.
These documents distill complex, scattered knowledge from YouTube tutorials and official docs into structured, actionable prompt engineering frameworks (e.g., 8-element formula, Kling 5-layer). They represent a reusable pattern for creating a DSL for prompt construction. An engineer could extract these formulas and apply them to a completely different set of models or generative domains.
A Claude Code Skill that converts prompts between different AI video models using the `cross-model-matrix.json` as a lookup table. This demonstrates a non-obvious pattern of using structured benchmark data to power an AI assistant skill. The `SKILL.md` format with YAML frontmatter is a reusable pattern for creating executable knowledge layers.