🕵️‍♂️ Sherlock AI Plugin

Paper to Comics

The Ultimate Research & Implementation Assistant for Claude Code & Cursor

Sherlock is a curated collection of high-powered AI skills designed to transform your research workflow. It doesn't just read papers—it deduces, visualizes, and reconstructs them into working code.

From deep-diving into literature to converting methodology into Python, Sherlock equips your AI agent with the tools of a master detective and a master engineer.

🧰 The Essentials Kit (Features)

Sherlock equips your agent with these specialized skills:

🧩 1. Paper2Code (The Engineer)

Transforms academic theory into executable reality.

• 4-Stage Pipeline: Systematically converts research papers into code (Algorithm Extraction → Concept Analysis → Planning → Implementation).
• No Hallucinations: Forces a structured intermediate representation (YAML) before writing a single line of code.
• Reproducibility First: Prioritizes accuracy and paper fidelity over "clever" coding.

📚 2. Deep Research (The Detective)

The heavy lifter for comprehensive reports.

• Multi-Pass Drafting: Spawns parallel sub-agents to draft different sections of a report.
• Evidence Tracking: Maintains strictly cited evidence tables—no claim goes unsourced.
• High Fidelity: Produces professional-grade reports with strict formatting compliance.

🎨 3. Paper Comic (The Storyteller)

Explains the unexplainable through visual narratives.

• Visual Translation: Turns dense academic text into educational comics.
• Style Adaptive: Choose from Classic, Tech/Futuristic, Warm, or Chalkboard art styles.
• Gemini Powered: Uses the genimg-gemini-web skill for consistent character consistency across panels.

🔬 4. Paper Analyzer (The Analyst)

X-Ray vision for PDFs.

• MinerU Integration: High-precision parsing of formulas, tables, and latex from PDFs.
• Style Rewrite: Can rewrite complex papers into "Storytelling", "Academic", or "Concise" formats.
• Metadata Extraction: Automatically pulls title, authors, and citations.

🖼️ 5. GenImg Gemini Web (The Artist)

The visual engine.

• Image Generation: Generates images via Google's Gemini Web.
• Multi-Modal: Handles text-to-image and image-to-text tasks.
• Session Awareness: Maintains context across multi-turn conversations for consistent output.

📐 6. Visual Architect (The Designer)

Blueprints for understanding.

• Schema Generation: Transforms methodology sections into structural visual schemas.
• Prompt Engineering: Generates high-precision prompts for DALL-E 3 or Midjourney based on paper logic.
• Layout Logic: Detects if a system is Linear, Cyclic, Hierarchical, or Parallel.

---

🚀 Quick Start

Installation

Option 1: CLI Install (Recommended)

Use npx skills to install skills directly:

# Install all skills
npx skills add proyecto26/sherlock-ai-plugin

# Install specific skills
npx skills add proyecto26/sherlock-ai-plugin --skill paper-comic genimg-gemini-web

# List available skills
npx skills add proyecto26/sherlock-ai-plugin --list

This automatically installs to your .claude/skills/ directory.

Option 2: Claude Code Plugin

Install via Claude Code's built-in plugin system:

# Add the marketplace
/plugin marketplace add proyecto26/sherlock-ai-plugin

# Install all marketing skills
/plugin install sherlock-ai-plugin

Option 3: Codex / Cursor / Copilot CLI (Agent Plugins)

This plugin follows the Agent Plugins 1.0 open standard, so the same repository installs on any conforming host:

Host	Install
OpenAI Codex	codex plugin marketplace add proyecto26/sherlock-ai-plugin then codex plugin add sherlock-ai-plugin@sherlock-ai-plugin-marketplace (or pick it from /plugins)
GitHub Copilot CLI	copilot plugin marketplace add proyecto26/sherlock-ai-plugin then copilot plugin install sherlock-ai-plugin@sherlock-ai-plugin-marketplace
Cursor	Settings → Plugins → install from Git URL https://github.com/proyecto26/sherlock-ai-plugin (Agent Plugins format is detected from the root plugin.json)

Which file each host reads:

• plugin.json (repo root) — the portable Agent Plugins 1.0 manifest. Cursor and Copilot CLI read this directly and discover skills/ automatically.
• .codex-plugin/plugin.json + .agents/plugins/marketplace.json — OpenAI Codex manifest and catalog.
• .claude-plugin/plugin.json + .claude-plugin/marketplace.json — Claude Code manifest and catalog (Copilot CLI also reads this marketplace file).

All six skills are discovered from skills/ — no host manifest repeats the list, so adding a skill is a one-directory change. Run bash scripts/check-manifests.sh before releasing to confirm every manifest still agrees.

Copilot CLI note: installing from a working tree that contains .git/ can fail with access denied while Copilot copies the directory. Install from a clean checkout or from the marketplace.

Option 4: Clone and Copy

Clone the entire repo and copy the skills folder:

git clone https://github.com/proyecto26/sherlock-ai-plugin.git
cp -r sherlock-ai-plugin/skills/* .claude/skills/

Option 5: Git Submodule

Add as a submodule for easy updates:

git submodule add https://github.com/proyecto26/sherlock-ai-plugin.git .claude/sherlock-ai-plugin

Then reference skills from .claude/sherlock-ai-plugin/skills/.

Option 6: Fork and Customize

1. Fork this repository
2. Customize skills for your specific needs
3. Clone your fork into your projects

Authentication

Some skills (like paper-analyzer) require API tokens or login sessions.

• MinerU: Export MINERU_TOKEN in your environment.
• Gemini: genimg-gemini-web needs a signed-in Google session. Run npx -y bun skills/genimg-gemini-web/scripts/main.ts --login and complete sign-in in the Chrome window that opens.

Image generation requires being signed in. Google serves signed-out visitors a free Flash-tier model that declines to create images. Cached cookies expire, so if image generation stops working, re-run --login.

Usage Examples

"Implement the code of this paper for me."

Triggers Paper2Code to read the PDF, plan the architecture, and write the Python implementation.

"Explain this complex transformer architecture as a comic."

Triggers Paper Comic to create a visual storyboard explaining the concept.

"I need a deep research report on the state of LLM reasoning."

Triggers Deep Research to crawl, analyze, and compile a multi-page cited report.

"Visualize the flow of data in this system."

Triggers Visual Architect to design a schematic diagram prompt.

---

📂 Structure

plugin.json              # Agent Plugins 1.0.0 manifest (Cursor, Copilot CLI)
.codex-plugin/           # OpenAI Codex manifest
.claude-plugin/          # Claude Code manifest + marketplace
.agents/plugins/         # Codex / Copilot catalog entry
scripts/
└── check-manifests.sh   # Asserts every host manifest agrees
skills/
├── deep-research/       # Report generation & evidence tracking
├── paper2code/          # Paper implementation pipeline
├── paper-comic/         # Educational comic generator
├── paper-analyzer/      # PDF parsing & style rewriting
├── genimg-gemini-web/   # Image generation backend
└── visual-architect/    # Visual schema design

---

🌟 Star History

💜 Sponsors

This project is free and open source. Sponsors help keep it maintained and growing.

Become a Sponsor | Sponsorship Program

🤝 Contribution

When contributing to this repository, please first discuss the change you wish to make via issue, email, or any other method with the owners of this repository before making a change.

Contributions are what make the open-source community such an amazing place to learn, inspire, and create. Any contributions you make are greatly appreciated ❤️.

You can learn more about how you can contribute to this project in the contribution guide.

👍 Credits

• Claude Code skills for academic papers

Happy vibe researching 💯

Made with ❤️ by Proyecto 26 - Changing the world with small contributions.

One hand can accomplish great things, but many can take you into space and beyond! 🌌

Together we do more, together we are more ❤️