edge_source_skill

AI-powered source intelligence recommender — find high-depth information sources for edge knowledge, adversarial SEO, and OSINT research.

中文说明

What It Does

Give it a research question, get back a scored, ranked list of information sources with reading order and predicted knowledge value.

python scripts/recommend_sources.py "I want adversarial SEO sources to understand edge mechanics"

Output:

• Query tags — auto-detected topic clusters
• Top sources — scored on Relevance × Depth × Actionability × Freshness × Consequence
• Reading order — fast execution → deep method → adversarial counterexample
• Knowledge gaps — missing source types to expand coverage
• 24h experiment — one actionable test you can run today

Pipeline

This is Stage 1 of a 3-stage content intelligence pipeline:

edge_source_skill ──→ EdgeKnowledge_Skill ──→ anything-to-md
   (find sources)       (crawl & analyze)        (convert to MD)

Stage	Repo	What it does
1 — Find	edge_source_skill	Recommend high-value sources for a topic
2 — Crawl	EdgeKnowledge_Skill	Deep-crawl forums, Reddit, communities
3 — Convert	anything-to-md	Convert any content to LLM-ready Markdown

Data Sources

File	Contents
references/karpathy-92-hn-2025.tsv	92 long-form high-signal feeds curated by Karpathy
references/deep-sources-github.tsv	Expanded deep-source GitHub projects (awesome lists, OSINT frameworks, security tools)
references/deep-sources-telegram.tsv	Telegram channel pool (continuously importable)

Included Deep GitHub Sources

awesome · awesome-osint · OSINT-Framework · awesome-security · awesome-incident-response · papers-we-love · awesome-selfhosted · awesome-sysadmin · awesome-privacy · awesome-threat-detection · awesome-browser-security · awesome-malware-analysis · awesome-devsecops

Scoring Formula

KnowledgeValue = 0.30×Relevance + 0.25×Depth + 0.25×Actionability + 0.10×Freshness + 0.10×Consequence

Each dimension scored 1–5. Sources ranked by weighted KnowledgeValue.

Telegram Integration (Optional)

pip install telethon

export TG_API_ID=your_id
export TG_API_HASH=your_hash

python scripts/import_telegram_sources.py --query "seo blackhat osint" --limit 30 --tags "telegram,seo,adversarial"

Imported channels auto-merge into the recommendation pool.

Quick Start

git clone https://github.com/1596941391qq/edge_source_skill.git
cd edge_source_skill
pip install -r requirements.txt  # or just run — minimal deps
python scripts/recommend_sources.py "your research question here"

Related Projects

• EdgeKnowledge_Skill — Stage 2: deep-crawl the sources this tool recommends
• anything-to-md — Stage 3: convert crawled content to clean Markdown
• bypass-anything — Anti-detection layer for browser automation
• nichedigger — Reddit-powered keyword mining for PSEO