SuperAPI AI Directory
Partner中文AI newsFavorites
Skillsknowledge-absorber

knowledge-absorber

General & other

View repository

Updated 10-16 06:52

About this skill

Usage

Organize links, documents, images, or code into a learning-ready, searchable, and further-questionable finished learning product, and output:

  • knowledge_card.md
  • knowledge_card.interactive.html

Only one command is needed:

python scripts/run_full_pipeline.py “<链接或本地文件路径>” [“<更多链接或文件路径>”]

Example:

python scripts/run_full_pipeline.py “https://example.com/article”
python scripts/run_full_pipeline.py “E:\资料\AI导论.pdf”
python scripts/run_full_pipeline.py “E:\资料\架构图.png”
python scripts/run_full_pipeline.py “E:\repo\main.py”
python scripts/run_full_pipeline.py “https://example.com/article” “E:\资料\补充说明.pdf”

This natural language phrasing is also supported:

python scripts/run_full_pipeline.py “学习这个 https://example.com/article”
Core Workflow

This skill uses the Truth Anchoring Protocol to ensure content accuracy.

Step 1: Intelligent Ingestion (Content Ingestion)

Run the script to obtain clean content. The script automatically cleans HTML noise, ads, and useless content, and handles multimodal content (PDF/OCR).

Cleaning Mechanism：

  • Automatically remove UI elements such as ads, navigation, footers, and social sharing
  • Filter out short texts containing noise keywords (login, register, download, follow, etc.)
  • Identify and remove ad containers (through class/id pattern matching)
  • Do not retain obvious UI elements and marketing content
  • Do not retain donation prompts such as "Thanks for the support, boss"
  • Do not retain page navigation and recommended links

Step 2: Truth Anchoring (Truth Anchoring)

"Do not trust any text lightly, even if it looks very professional."

Before generating content, the ingested content must first undergoaccuracy verification。

Verification Process：

  1. Extract core claims: Identify all key factual claims in the text (specific data, API usage, historical events, absolute assertions)
  2. Online audit: Call WebSearch to verify each claim; searches must include the current year (2026) to ensure timeliness
  3. Generate a calibration report: Build a "red team report" marking erroneous, outdated, or disputed content
  4. Explicit labeling: Clearly label all uncertain or disputed information in the final output

Verification Requirements：

  • Every key claim must be verified through search
  • Unverified original content must not be used directly
  • Obvious timeliness issues must not be ignored (such as "latest in 2020")
  • Verification of absolute assertions must not be omitted ("always", "never", "must")
  • After discovering an error, it must not be used directly without labeling

Labeling Standards：

  • Outdated information must be labeled "[Outdated]"
  • Disputed content must be labeled "[Disputed]"
  • Content that cannot be confirmed must be labeled "[To Be Confirmed]"

Step 3: Generate Teaching Content

Based on the cleaned content and calibration report, generate multimodal output.

Content Adaptation：

  • Technical: Use a modern, clean style (Apple Design style)
  • Traditional Chinese studies/humanities: Use an ink-wash traditional Chinese studies style (traditional Chinese fonts, ink-wash background, seal decorations)

Traditional Chinese studies style trigger conditions: When the title, source, tags, or the first few paragraphs of the content contain any of the following keywords, automatically apply the traditional Chinese studies style:

  • Classic Literature: The Analects, Zhuangzi, Tao Te Ching, Records of the Grand Historian, Classic of Poetry, I Ching, Mencius, Xunzi, Spring and Autumn Annals, Zuo Zhuan, Book of Rites, Zhou Yi, Book of Documents, Great Learning, Doctrine of the Mean
  • School of Thought: Chinese studies, classical Chinese, Buddhism, Confucianism, Daoism, philosophy, humanities, Confucian school, Daoist school, Buddhist school, Legalist school, Mohist school, Military school, Chan Buddhism, Neo-Confucianism, School of Mind, Xuanxue
  • Historical Figures: Confucius, Laozi, Mencius, Zhuangzi, Xunzi, Mozi, Han Feizi, Zhu Xi, Wang Yangming, Cheng Yi, Cheng Hao
  • Literary Genres: ancient poetry, ci, fu, parallel prose, prose, classical Chinese
  • Other: classics, histories, masters, and collections; the Four Books and Five Classics; the Hundred Schools of Thought; traditional culture

Step 4: Quality Assurance

The following items must be confirmed before delivery:

  1. ✅ HTML includes <script> search logic
  2. ✅ In Chinese studies mode, use traditional Chinese fonts (Songti, Source Han Serif) and ink-wash style
  3. ✅ Includes a Mermaid cognitive map
  4. ✅ Includes 5-8 FAQs
  5. ✅ Content has been cleaned up (no ads, no UI elements, no donation prompts)
  6. ✅ All key claims have been verified through truth anchoring
  7. ✅ Uncertain content is clearly marked

The following situations must not occur：

  • Outputting HTML containing ads or marketing content
  • Outputting unverified factual claims
  • Using modern Western fonts in Chinese studies mode
  • Omitting the search function
  • Omitting Mermaid diagrams

If any item does not meet the requirements, it must be regenerated.

Output Location and Files

The script will generate an output folder under outputs/knowledge_YYYYMMDD_* , and in the end only two files will be kept:

  • knowledge_card.md
  • knowledge_card.interactive.html
Tutor Mode (interactive HTML optional)

Tutor mode requires a local relay to call the model in the browser:

  • python scripts/mentor_relay.py(default http://127.0.0.1:8760）

If you are only reading/searching the main text, you do not need to start the relay.

Supported Inputs
  • URL links
  • PDF / Word / Markdown / TXT
  • PNG / JPG / JPEG / WEBP
  • Code files such as Python / JavaScript / TypeScript / Java / C / C++ / Go / Rust
  • Pass in multiple links or files at once