Survey Generator

Survey Generator is a Claude Code plugin from DAIR.AI Academy. Give it an AI or machine-learning topic and a public source to start from, and it produces an academic-style survey paper as one self-contained HTML file, with numbered sections, inline SVG figures and a References list. Your agent does the research curation; Kimi K2.6 on Fireworks AI writes the paper in a single call, so you need a Fireworks API key. What you can do with it - Survey a topic such as reasoning models or agentic engineering, starting from an arXiv survey, a curated GitHub list or a well-known blog post - Choose the size: about 20 papers for a quick survey, 40 to 50 for a comprehensive one, or 80 to 100 for an exhaustive one - Start from DAIR.AI's AI Papers of the Week, an open index of notable papers, for broad topics - Rerun with a sharper style spec or research bundle; each run writes a new versioned file, so you can compare them How it works - Your agent reads the source, drafts a taxonomy of 4 to 8 branches and 6 to 10 sections (introduction, foundations, methods, evaluation and open problems), and curates a bibliography of real papers, each with its authors, year, title, venue and a short summary. - It saves that as research_bundle.json and runs build_artifact.py, which sends it with a fixed style spec to Kimi K2.6 on Fireworks. - The survey is written to output/survey_kimi-k2p6_v<N>.html in the skill's folder. Open it in a browser. - Its rules: never invent papers, and change the inputs rather than editing the generated HTML. What it needs - A Fireworks AI account, with FIREWORKS_API_KEY set in your environment. The model calls run on your Fireworks account. - Python 3, with no extra packages. - Set FIREWORKS_MODEL to try another Fireworks model, such as Kimi K2.5. How to install it In Claude Code, run /plugin marketplace add dair-ai/dair-academy-plugins, then /plugin install survey-generator@dair-academy-plugins. Good to know - Your research bundle (the topic, taxonomy and bibliography) is sent to Fireworks AI. - A worked 100-paper example on agentic engineering comes with the skill.

• “Run the survey generator on "Reasoning Models" using AI Papers of the Week as the anchor source. Target 50 papers.”

  Copy

• “Write a survey paper on diffusion language models, starting from this arXiv survey.”

  Copy

• “Generate a quick 20-paper survey on agentic engineering.”

  Copy

• “Make a literature review artifact on retrieval-augmented generation from this awesome list.”

  Copy