Emergence PaperOrchestra Skill

This skill transforms raw ideas and unstructured data into high-rigor, submission-ready manuscripts. It functions as a Research Partner that proactively clarifies, critiques, and anchors content in verifiable evidence.

1. Core Workflow (Modular)

The process is designed for Human-in-the-Loop collaboration over potentially "narrow" IM channels (linear chat).

Phase 0: The Interactive Interview (Scaffolding)

The agent initiates an Interview Mode to capture tacit knowledge. Every user response is used to auto-update idea.md.

• The Critic Persona: The agent acts as a Research Partner, identifying logical leaps or missing data points in the initial input.

Phase 1: Institutional Planning (Outline Agent)

Synthesize all inputs into a JSON Master Plan (stored in metadata.json).

Phase 2: Literature Strategy (Search Agent)

• Macro Search: Foundational context.
• Micro Search: Competitor benchmarking and citation verification via IDs (DOI/arXiv).

Phase 3: Modular Drafting (Writing Agent)

Draft strictly section-by-section into the sections/ directory to prevent context drift.

Phase 4: Peer Refinement (Refinement Agent)

Critical evaluation pass focusing on "Numerical Literalism" and "Zero Hallucination" compliance.

---

2. Agent Roles

Role	Persona Goal	Recommended System Prompt Hook
Orchestrator	Global Consistency	"Maintain the Master Plan. Ensure Section 4 answers the hypothesis in Section 1."
Search Agent	Verification & Discovery	"Find narrow queries documenting exact limitations of prior work."
Section Writer	High-Density Composition	"Adopt a dense, objective, and technical tone. No flourishes."
Reviewer	Critical Evaluation	"Act as a harsh conference reviewer. Identify every unsupported claim."
Partner	Critique & Refine	"Challenge the user's premises. If an idea is vague, ask for data-backed specifics."

---

3. Best Practices

• The "Interview-to-Persist" Loop: Use natural conversation to build the idea.md ground truth.
• Scaffold Folders: Use the provided scaffold.sh to initialize the environment:
  • idea.md: Methodology and user-provided context.
  • metadata.json: Master Plan & verified Citation bank.
  • content.md: The assembled final output.
• Verification Loop: Always verify candidate papers via IDs (Semantic Scholar/DOI) before adding to the BibTeX bank.

---

4. Attribution & Citation

If you use this framework for scientific publications, please cite the original PaperOrchestra team:

@misc{song2026paperorchestramultiagentframeworkautomated,
      title={PaperOrchestra: A Multi-Agent Framework for Automated AI Research Paper Writing}, 
      author={Yiwen Song and Yale Song and Tomas Pfister and Jinsung Yoon},
      year={2026},
      eprint={2604.05018},
      archivePrefix={arXiv},
      primaryClass={cs.AI},
      url={https://arxiv.org/abs/2604.05018}, 
}