Add Ayanami0730/paper-read-skill

agents/Ayanami0730__paper-read-skill/README.md

@@ -0,0 +1,56 @@
+# paper-read-skill
+
+An AI agent skill-set for **deep, rigorous reading of academic research papers and technical blog posts**. Drop it into Cursor, Claude Code, Codex, or OpenCode — then hand it a paper URL, PDF, DOI, or blog link and get a rich, exhaustive analysis in Chinese prose.
+
+## What it does
+
+The agent auto-routes to a type-specific analysis framework based on what you give it:
+
+### 📄 Research Papers (`paper-analysis` skill)
+
+| Paper type | What the agent does |
+|---|---|
+| **Benchmark** | Dissects dataset construction, evaluation protocols, metrics (with LaTeX formulas), reproducibility risk |
+| **Methodology / Algorithm** | Unpacks architecture, training strategy, technical innovation; situates in the research timeline |
+| **Survey / Opinion** | Audits argument structure, coverage gaps, author biases |
+
+For every paper the agent:
+- Traces the **full citation network** — direct baselines are mandatory, contemporaneous work is searched
+- **Profiles core contributors** — first/corresponding author academic trajectory, prior work, Twitter/X discussion
+- Reports **community signals** — citation growth, GitHub stars, endorsements from prominent researchers
+- Explains every **evaluation metric** with a LaTeX formula and a concrete worked example
+- Checks **GitHub repos, Hugging Face pages, and issue/discussion threads** for real-world adoption data
+
+### 📝 Tech Blogs (`blog-reading` skill)
+
+| Blog type | What the agent does |
+|---|---|
+| **Technical-research** | Explains engineering principles, assesses innovation depth and evidence strength |
+| **Product-launch** | Cuts through marketing language, extracts genuine technical progress and user value |
+| **Industry-opinion** | Deconstructs argument logic, surfaces unstated assumptions, presents countervoices |
+| **Case-story** | Separates PR polish from real field experience, surfaces transferable patterns |
+
+Output is **fluent Chinese prose** — no markdown bullet lists in the main body, coherent paragraphs with logical transitions, as if narrated by a senior technical editor to a peer.
+
+## Example usage
+
+```
+# In Claude Code or Cursor:
+Read this paper: https://arxiv.org/abs/2305.10601
+```
+
+```
+Analyze this blog post: https://cursor.sh/blog/shadow-workspace
+```
+
+## Install
+
+```bash
+npx skills add Ayanami0730/paper-read-skill -g
+```
+
+Or copy `skills/` subdirectories manually to `~/.cursor/skills/`.
+
+## Compatible runtimes
+
+Cursor · Claude Code · OpenAI Codex · OpenCode

agents/Ayanami0730__paper-read-skill/metadata.json

@@ -0,0 +1,27 @@
+{
+  "name": "paper-read-skill",
+  "author": "Ayanami0730",
+  "description": "Deep analysis agent for academic papers and tech blogs. Auto-routes to type-specific frameworks for benchmark, methodology, survey/opinion papers and four blog categories.",
+  "repository": "https://github.com/Ayanami0730/paper-read-skill",
+  "version": "1.0.0",
+  "category": "research",
+  "tags": [
+    "research",
+    "papers",
+    "academic",
+    "analysis",
+    "blog-reading",
+    "chinese",
+    "claude-code",
+    "cursor"
+  ],
+  "license": "MIT",
+  "model": "claude-sonnet-4-6",
+  "adapters": [
+    "claude-code",
+    "cursor",
+    "openai",
+    "system-prompt"
+  ],
+  "icon": false
+}