Outline Agent

This is the first stage of the PaperOrchestra pipeline from arXiv:2604.05018, a research paper workflow system. It reads your raw materials (idea document, experimental logs, LaTeX template, conference guidelines) and produces a strict JSON outline with three components: a plotting plan that specifies figure types and data sources, a literature search strategy that separates intro breadth from related work depth, and a section-by-section writing plan with mandatory citation hints for every dataset and model mentioned. The implementation is faithful to the paper's appendix, down to enforcing timeline cutoffs so you don't cite work published after your conference deadline. Validation is strict because downstream agents depend on the schema. Use this when you want structured planning before writing, not freeform drafting.

GitHub stars: 572
last updated: Jun 2, 2026

View on GitHub

Install

$ npx -y skills add ar9av/paperorchestra --skill outline-agent --agent claude-code

Installs into .claude/skills of the current project.

Files

Outline Agent (Step 1)

Faithful implementation of the Outline Agent from PaperOrchestra (Song et al., 2026, arXiv:2604.05018, App. F.1, pp. 40–44).

Cost: 1 LLM call.

Your task

Read four input files from the workspace and produce a single JSON object at workspace/outline.json with three top-level keys:

plotting_plan — array of figure objects
intro_related_work_plan — object with introduction_strategy and related_work_strategy
section_plan — array of section objects, each with section_title and subsections[]