anime-video-mcp
# One-minute anime knowledge video skill
A Codex Agent Skill and local MCP server for **planning** cited blue-white anime knowledge shorts and an original continuing romance series. The female and male masters supplied with this project are in `references/`.
**Current status:** installable Python package; offline planning for one curated astronomy topic and three original romance chapter fixtures; no production MP4 pipeline yet. Media tools intentionally return blocking status. See [中文说明](README_zh-CN.md) for setup and limitations.
Quick check with Python 3.10+:
```bash
python -m pip install -e '.[mcp]'
anime-video episode '为什么夜空是黑的？' --series cosmos --quality draft
anime-video romance-series
anime-video romance-chapter 1
anime-video-mcp
```
The MCP process speaks stdio and does not print a startup banner. The official SDK is preferred; a minimal dependency-free stdio implementation supports offline planning when PyPI is unavailable. The skill is at `skills/one-minute-anime-knowledge/SKILL.md`.
TDQS
Scored across 12 tools
Most tools have clearly distinct names and purposes, but five tools (generate_assets, generate_voice, generate_music, render_episode, qa_episode) share identical descriptions ('Report pending media provider or QA gate'), which could cause misselection if an agent relies solely on descriptions. The distinct names mitigate this, but the overlap in descriptions introduces some ambiguity.
All tool names consistently follow a verb_noun pattern (e.g., create_episode, build_storyboard, generate_voice, qa_episode), with no mixed conventions or style deviations. This makes the naming predictable and easy to navigate.
With 12 tools, the count falls comfortably within the ideal 3-15 range. Each tool represents a distinct step in the anime production pipeline (research, script, storyboard, asset/voice/music generation, render, QA, and batch/series creation), so no tool feels redundant or missing.
The pipeline covers the full lifecycle from research to QA, including script, storyboard, and media generation. Minor gaps exist, such as the absence of explicit update/delete operations or a tool to list existing episodes, but these are not critical for the domain and agents can likely work around them.