ar9av/paperorchestra
9 skills415 installs5.7k starsGitHub
Install
npx skills add https://github.com/ar9av/paperorchestraSkills in this repo
1Paper OrchestraThe top-level orchestrator for the PaperOrchestra pipeline that turns unstructured research materials into a submission-ready LaTeX manuscript and compiled PDF. It coordinates five specialized agents (outline, plotting, literature-review, section-writing, content-refinement) and runs pre-flight validation and anti-leakage guards. A developer runs it to write a full conference paper from an idea and experimental log.58installs2Paper AutoratersRuns the four paper-quality autoraters from PaperOrchestra: Citation F1 with a P0/P1 partition, a 6-axis 0-100 Literature Review Quality score with anti-inflation rules, and two side-by-side LLM-as-judge comparators for overall paper and literature-review quality. A developer uses it to score a generated paper against ground truth or compare two paper-writing pipelines.46installs3Plotting AgentStep 2 of the PaperOrchestra pipeline that executes the visualization plan from the outline, rendering plots and conceptual diagrams from the experimental log and idea files. It optionally refines each figure through a vision-model critique loop and produces context-aware captions. A developer runs it to generate the figures a paper needs, in parallel with the literature-review agent.46installs4Content Refinement AgentStep 5 of the PaperOrchestra pipeline that iteratively refines a paper draft by simulating peer review and applying targeted revisions. It uses deterministic accept/revert halt rules, per-iteration snapshots, and a hallucination check against the experimental log so acceptance only happens on genuine score improvement. A developer runs it to polish a generated LaTeX manuscript before submission.45installs5Agent Research AggregatorA pre-pipeline aggregator that scans AI coding-agent cache directories for scattered experimentation logs, extracts insights and numeric results, and formats them into the structured idea.md and experimental_log.md inputs the PaperOrchestra pipeline expects. A developer runs it before paper-orchestra when they only have raw agent history rather than clean inputs.44installs6Literature Review AgentStep 3 of the PaperOrchestra pipeline that discovers candidate papers via web search, verifies them through Semantic Scholar with fuzzy title matching and temporal cutoffs, and cross-corroborates against Crossref and OpenAlex to catch hallucinated citations. It then builds a BibTeX file and drafts the Introduction and Related Work sections. A developer runs it to assemble a grounded bibliography for a generated paper.44installs7Outline AgentStep 1 of the PaperOrchestra pipeline that converts raw materials (idea, experimental log, LaTeX template, conference guidelines) into a strict JSON outline. The outline contains a plotting plan, a literature-search plan for the Introduction and Related Work, and a section-level writing plan with citation hints. A developer runs it first because every downstream agent parses this JSON.44installs8Paper Writing BenchReverse-engineers raw materials (a Sparse idea, a Dense idea, and an experimental log) from an existing research paper to build a benchmark case for evaluating paper-writing pipelines. It replicates the PaperWritingBench dataset-construction procedure by stripping narrative flow from a paper PDF. A developer uses it to create test inputs, run their pipeline, and compare output against the original paper.44installs9Section Writing AgentStep 4 of the PaperOrchestra pipeline: a single multimodal LLM call that drafts the remaining paper sections (Abstract, Methodology, Experiments, Conclusion), extracts numeric values from the experimental log into LaTeX booktabs tables, splices in the generated figures, and merges everything into the template that already holds the Introduction and Related Work. A developer runs it to fill in the body of a paper.44installs