Skip to content

Latest commit

 

History

24 Commits

Folders and files

NameName
Last commit message
Last commit date
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

A literature survey is not reading — it is systematic cartography. You are mapping terrain that no one has drawn before, and the quality of your map determines whether the expedition succeeds or fails.

📚 Literature Survey

Autonomous Literature Survey Engine for Academic Research

Literature Survey is not a search assistant that finds papers for you. It is the surveyor. You provide a research question — it searches, screens, reads, categorizes, identifies gaps, and produces a structured survey output. Autonomously. Five research paradigms, each a self-contained playbook with quantitative budget enforcement.


⚡ What It Does

  • 🗺️ Scoping survey — broad landscape mapping of a new field. High volume (100 papers), shallow depth. Produces a hierarchical field taxonomy with key authors, trends, and open questions
  • 📋 Systematic survey — exhaustive PRISMA-style coverage. Multi-stage screening, citation chaining, quality assessment. Produces comparison tables, PRISMA flow, and gap analysis
  • 🔬 Deep survey — precise investigation of a specific sub-problem. Few papers, all read in full (50% deep-read rate). Produces detailed technical analysis with equations and hyperparameters
  • 📝 Narrative review — theory-driven argument building. Strategic evidence selection, thematic coding. Produces a structured narrative with thesis and counter-evidence
  • 🔗 Snowball — citation-chain-driven exploration from seed papers. Forward and backward tracing until saturation (67% deep-read rate). Produces a research lineage map

🎯 Design Philosophy

🎖️ Strategy-Book Pattern

This is not a chatbot. After brief intake (if needed), CC selects a strategy and executes autonomously. Each strategy is a self-contained playbook with:

  • A quantitative budget (hard floors, ±10% flexibility)
  • Available tactics and SOPs (CC decides sequencing)
  • Quality gates (inherited from imported base skills)
  • A defined output format

The human provides only the research question. Everything after that — what to search, which papers to read deeply, when to stop, how to organize findings — is decided by the engine.

📐 Three-Layer Command Structure

Strategy (5 available)
  → Budget + available tactics + execution guidance
  → CC decides: which tactics, what order, how many iterations

Tactic (3 available)
  → Multi-step orchestration pattern
  → Composes multiple SOPs into coherent workflows

SOP (17 available: 5 import + 11 subagent + 1 dialogue)
  → Single focused operation
  → Import: delegates to external skill
  → Subagent: isolated context execution
  → Dialogue: talks to user (intake only)

📏 Budget Enforcement

AI agents naturally take the path of least resistance — searching a handful of papers and declaring victory. Every strategy embeds hard enforcement:

  • Quantitative floors: Each SOP has a target count per strategy. CC cannot exit until ≥90% of each target is met.
  • State Ledger: A progress table printed before each iteration decision — the AI cannot lose track of where it stands.
  • Deviation documentation: Any departure from budget requires explicit reasoning.

🏗️ Architecture

┌───────────────────────────────────────────────────────────────┐
│  ENTRY POINT (/literature-survey)                             │
│  Routes to strategy based on user intent                      │
├───────────────────────────────────────────────────────────────┤
│  STRATEGY (5)                                                 │
│  scoping-survey, systematic-survey, deep-survey,              │
│  narrative-review, snowball                                   │
├───────────────────────────────────────────────────────────────┤
│  TACTIC (3)                                                   │
│  prisma-screening, citation-chaining, narrative-framing       │
├───────────────────────────────────────────────────────────────┤
│  SOP (17)                                                     │
│  import (5) | subagent (11) | dialogue (1)                    │
├───────────────────────────────────────────────────────────────┤
│  EXTERNAL SKILLS + MCP TOOLS                                  │
│  web-browsing, literature-engine, subagent-spawning,          │
│  brave-search, apify, alphaxiv, semantic-scholar              │
└───────────────────────────────────────────────────────────────┘

🧩 Skill Inventory

Category Count Skills
Strategies 5 scoping-survey, systematic-survey, deep-survey, narrative-review, snowball
Tactics 3 prisma-screening, citation-chaining, narrative-framing
Import SOPs 5 web-search, web-research, paper-overview, paper-search, paper-research
Subagent SOPs 11 survey-synthesis, define-search-protocol, categorize-papers, extract-data, quality-assessment, seed-selection, saturation-detection, taxonomy-mapping, prisma-flowchart, thematic-coding, gap-identification
Dialogue SOPs 1 intake-dialogue

📊 Budget Table

Strategy web-search web-research paper-overview paper-search paper-research
scoping-survey 100 10 100 20 0
systematic-survey 50 5 60 40 30
deep-survey 30 5 40 40 20
narrative-review 80 15 50 40 20
snowball 20 3 30 30 20

All values are defaults with ±10% flexibility. CC documents deviations with reasoning.

📁 Repository Structure

literature-survey/
├── skills/
│   ├── literature-survey/        # Entry point — /literature-survey
│   ├── scoping-survey/           # Strategy
│   ├── systematic-survey/        # Strategy
│   ├── deep-survey/              # Strategy
│   ├── narrative-review/         # Strategy
│   ├── snowball/                 # Strategy
│   ├── prisma-screening/         # Tactic
│   ├── citation-chaining/        # Tactic
│   ├── narrative-framing/        # Tactic
│   ├── web-search/               # Import SOP
│   ├── web-research/             # Import SOP
│   ├── paper-overview/           # Import SOP
│   ├── paper-search/             # Import SOP
│   ├── paper-research/           # Import SOP
│   ├── survey-synthesis/         # Subagent SOP
│   ├── define-search-protocol/   # Subagent SOP
│   ├── categorize-papers/        # Subagent SOP
│   ├── extract-data/             # Subagent SOP
│   ├── quality-assessment/       # Subagent SOP
│   ├── seed-selection/           # Subagent SOP
│   ├── saturation-detection/     # Subagent SOP
│   ├── taxonomy-mapping/         # Subagent SOP
│   ├── prisma-flowchart/         # Subagent SOP
│   ├── thematic-coding/          # Subagent SOP
│   ├── gap-identification/       # Subagent SOP
│   └── intake-dialogue/          # Dialogue SOP
├── tests/
│   └── integration-prompt.md
└── README.md

🔌 Dependencies

Dependency Repository What It Provides
web-browsing yogsoth-ai/web-browsing web-search + web-research
literature-engine yogsoth-ai/literature-engine literature-overview + literature-search + literature-research
subagent-spawning yogsoth-ai/subagent-spawning Subagent dispatch conventions

🔧 MCP Servers Required

Server Purpose
brave-search Web search API
apify Google Scholar scraping + full page reading
alphaxiv Paper search, content extraction, PDF queries
semantic-scholar Paper lookup, citations, references, recommendations

🚀 Quick Start

  1. Clone:
git clone https://github.com/yogsoth-ai/literature-survey.git
  1. Ensure sibling skill repos are available: web-browsing, literature-engine, subagent-spawning

  2. Configure MCP servers (brave-search, apify, alphaxiv, semantic-scholar) in your Claude Code session.

  3. Invoke:

/literature-survey I need to understand all methods for protein structure prediction published since AlphaFold2

📄 License

Apache-2.0


Part of the Yogsoth AI ecosystem. Built by Pthahnix.

About

Autonomous Literature Survey Engine — 5 research paradigms (scoping, systematic, deep, narrative, snowball) for systematic academic investigation. Strategy-book pattern: CC reads and decides.

Topics

Resources

Stars

2 stars

Watchers

0 watching

Forks

Releases

Packages

Contributors