Open-source research pipeline

Literature review, novelty generation, citation verification, and adversarial review. A self-coordinating pipeline of 20+ specialized agents for researchers.

View on GitHub Novelty Skills

Quick Start

Python CLI tools:

$ git clone https://github.com/argahv/sisyphus-academica.git
$ cd sisyphus-academica && pip install -e .
$ sisyphus demo
Running pipeline on example topic: 'Transformer Efficiency' ...

Install skills via npx (any agent):

$ npx skills add argahv/sisyphus-academica -s contrarian
$ npx skills add argahv/novelty-skills -s heretic
Installed to Claude Code, Cursor, Copilot, Hermes Agent, OpenCode

Full pipeline (requires OpenCode):

$ git clone https://github.com/argahv/sisyphus-academica.git && cd sisyphus-academica
$ bash install.sh
25 agents deployed → select "research-director" → "write a paper about [topic]"

What Makes This Different

CapabilityOther AI Paper ToolsSisyphus Academica
Literature review10-20 papers500+ papers via 5 parallel scouts
Citation accuracy~60% (40% hallucination)100% verified against 2+ sources [view proof]
AI-sounding textPost-hoc cleanup41 Humanizer patterns as generation constraints
Voice calibrationNoneLearns author's voice from writing samples
Novelty generation"What's the gap?"6 novelty engines × 50+ hypotheses
Adversarial reviewNone10 distinct reviewer personas
SIREN paper 13-page PDF · 26 verified citations · 3 figures · 0 hallucinated references · download PDF

What You Get

17 Portable Agent Skills

Install any skill in one command — works with Claude Code, Codex, Cursor, Gemini CLI, OpenCode:

npx skills add argahv/sisyphus-academica     # Browse all 17 interactively
npx skills add argahv/novelty-skills           # Standalone thinking tools pack

Invoke directly in your agent:

/contrarian "The claim: 'Attention is all you need'"
/heretic "Paper: 'Scaling Laws for Neural Language Models'"

6 Novelty Engines

C

Contrarian

Invert every well-established claim

Generates 10 counter-hypotheses by inverting claims across 6 axes: polarity, direction, scope, relevance, existence, priority.

X

Cross-Pollinator

Solutions from 15 distant fields

Maps concepts from astrodynamics onto biology, from monetary policy onto ML. Extract the mechanism, not the metaphor.

A

Assumption Excavator

Find what everyone assumed

Surfaces resource, behavioral, environmental, temporal, and causal assumptions. Tests what breaks if each is false.

G

Counterfactual Generator

The field without its key papers

Removes the most-cited papers from history. Traces ripple effects. Finds suppressed alternatives that deserve a second look.

P

Paradox Sifter

Contradictions everyone ignores

Cross-references Limitations sections across papers. Finds direct contradictions, mutual ignorance, hidden dependencies, and escalations.

H

Heretic

50 wild hypotheses from any abstract

The crown jewel. Generates 50 hypotheses from title+abstract alone, scores against the actual paper, finds the haunting idea.

The Quality Gates

✓
Citation Verification — Every reference checked against Semantic Scholar + CrossRef. Must be found in 2+ sources. No exceptions.
✓
Statistical Audit — Every p-value, effect size, sample size, and test selection validated. No p-hacking, no multiple comparison errors.
✓
AI-Pattern Detection — 41 Humanizer patterns scanned (30 base + 11 academic). Density must be < 2 violations per 1000 words.
✓
Style Audit — Zero em dashes. Voice must match the author's writing profile.
✓
Adversarial Review — All 10 reviewer personas must recommend acceptance. Not a subset. All 10.

Pipeline Overview

┌──────────────────────────────┐ │ Research Director │ │ (orchestrator) │ └────────┬─────────────────────┘ │ ┌────────────────────┼─────────────────────────┐ ▼ ▼ ▼ ┌───────────┐ ┌────────────────┐ ┌──────────────────────┐ │ Literature│ │ 6 Novelty │ │ Gap Analyzer │ │ Scout ×5 │ │ Engines │ │ + Methodology │ └───────────┘ └────────────────┘ └──────────────────────┘ │ ┌────────────────────┼─────────────────────────┐ ▼ ▼ ▼ ┌─────────────────────────────────────────────────────────────┐ │ Parallel Writing Swarm (5 agents + 41 Humanizer patterns) │ └──────────────────────────┬──────────────────────────────────┘ │ ┌─────────────────────────────────────────────────────────────┐ │ Verifier: Citations × Statistics × AI-Pattern Detection │ └──────────────────────────┬──────────────────────────────────┘ │ ┌─────────────────────────────────────────────────────────────┐ │ Adversarial Review: 10 reviewer personas (parallel) │ └──────────────────────────┬──────────────────────────────────┘ │ ┌─────────────────────────────────────────────────────────────┐ │ Style Auditor: Humanizer certification, em dash zero check │ └──────────────────────────┬──────────────────────────────────┘ │ ┌─────────────────────────────────────────────────────────────┐ │ Formatter: LaTeX template, BibTeX, figures, PDF │ └──────────────┬───────────────────────────┬──────────────────┘ │ │ ▼ ▼ ┌──────────┐ ┌──────────────┐ │ Paper │ │ Citation │ │ PDF │ │ Report │ └──────────┘ └──────────────┘

Project Structure

sisyphus-academica/
├── orchestrator/          # Research Director agent
├── subagents/             # Writing pipeline agents (writer, verifier, etc.)
├── novelty-engines/       # 6 novelty generation agents
├── reviewers/             # 10 adversarial reviewer personas
├── skills/                # 17 portable skill files (SKILL.md format)
├── tools/                 # Python CLI toolchain
├── templates/             # LaTeX venue templates
├── config/                # Agent configuration
├── examples/siren-paper/  # Full pipeline output (13-page paper)
├── tests/                 # Python unit tests
└── docs/                  # This documentation site
Installation Guide Architecture Novelty Engines