使用方法
- 复制安装命令,在自己的智能体环境安装前先检查源码包。
- 从原始文档核对所需工具、凭据和支持的宿主。
- 说明输入材料与期望交付物,并对照来源检查输出。
npx skills add https://github.com/nphu0811/pmg-skills --skill pmg-visual-assets开始前需了解
- 安装不包含模型订阅、API 额度或访问文件的权限。
- 仓库星数和安装量反映活跃情况,不等于独立质量或安全测试。
按源仓库内容呈现,保留标题、案例、代码、表格、链接以及原文引用的演示图片。
PMG Visual Assets Engine
pmg-visual-assets provides source-first media research, licensing assessment, and asset preparation for PowerPoint Motion Graphics. It is integrated with the canonical PMG engine via pmg.visual_assets.VisualAssetEngine.
Key Principles & Boundaries
- Search vs. Generation Distinction:
- Searching for existing authentic assets and generating synthetic assets are distinct operations.
- The engine prefers authentic permitted sources (Wikimedia Commons, NASA media archive, open vector repositories).
- Generative fallback is activated only when no suitable asset exists and a supported generator adapter is configured.
- Never fabricate search results, URLs, licenses, or generative output.
- Licensing & Rights Management:
- Every asset records provenance (
retrievedvsgenerated), license name, license URL, author attribution, and restrictions. - Trademarks and brand logos require explicit rights review.
- Evaluation Criteria:
- Semantic relevance to the presentation topic.
- Technical quality: resolution, aspect ratio, contrast, transparency.
- Suitability for downstream decomposition and native PowerPoint animation.
- Engine Integration:
- Prepares media files and structured
AssetRecordobjects. - Does not compile PowerPoint animations directly; hands off assets to
pmg-image-decomposer,pmg-core, andpmg-morph.
Python API Usage
from pmg.visual_assets import VisualAssetEngine, AssetRequirement, WikimediaProvider, NASAProvider
# 1. Initialize engine with configured providers
engine = VisualAssetEngine(providers=[WikimediaProvider(), NASAProvider()])
# 2. Define asset requirements
req = AssetRequirement(
query="Space exploration Saturn rocket",
kind="image",
min_width=1280,
min_height=720,
transparency=False,
)
# 3. Research and rank candidates
search_result = engine.research(req, limit=5)
# 4. Acquire permitted asset with download and quality validation
acquired = engine.acquire(req, output_dir="internal_qa/assets")
if acquired["status"] == "READY":
asset_record = acquired["asset"]
print(f"Acquired: {asset_record.title} at {asset_record.local_path}")
