What problem does it solve? Writing an academic-style survey of a fast-moving AI/ML topic requires reading dozens of papers, building a taxonomy, curating a real bibliography, and producing polished figures and layout. This Skill automates that pipeline: the agent curates the research, and a single Fireworks API call to Kimi K2.6 renders the complete survey artifact. ## Core Features & Use Cases - Structured research bundle: Builds a research_bundle.json containing title, taxonomy, paradigms, stack layers, sections, a comparison table, and a bibliography of real papers with summaries. - One-call HTML generation: build_artifact.py sends the bundle plus a topic-agnostic style_spec.json to Kimi K2.6 via the Fireworks chat completions API and writes a versioned, self-contained HTML file with inline SVG figures and a numbered References section. - Deterministic figure contracts: The style spec pins exact SVG geometry for the taxonomy tree, paradigm panels, and stack diagram, with hard invariants that prevent overlapping nodes and clipped labels. - Use Case: Ask for a survey on "Reasoning Models" anchored to the DAIR.AI Papers of the Week repo; the agent curates 40-50 real papers, writes the bundle, runs the script, and produces output/survey_kimi-k2p6_v1.html ready to open or host. ## Quick Start Use the survey-generator skill to create a survey on Agentic Engineering anchored to the DAIR.AI AI Papers of the Week repository with a 40-paper bibliography.
