A workflow: lay out the conceptual map (space) and the evolutionary lineage (time) simultaneously, letting you build three-dimensional cognition of an unfamiliar field in the shortest possible time. Looking at the map alone, you don't know which concepts are important; looking at the lineage alone, you don't know what the current technological landscape looks like. Only by crossing the two can you locate where you are and where you should enter from.
First read references/sci-cli.md—CLI detection steps, usage, and iron rules.
Then read references/output-system.md—HTML shell specification and design tokens.
Steps (fixed order)
Step 0 — Confirm the CLI is available
Run sci --version. If not installed, give the installation command and stop; if there is an authentication error, prompt sci init and stop.
Step 1 — Anchor the field, perform multidimensional retrieval (conceptual map side)
Take the input term as the central concept, perform the main retrieval (--mode low，--limit 20), then do at least 2 supplementary retrievals (method neighbors, application scenarios, contrast concepts, and other angles). From the abstracts, mine 5–8 neighboring concepts, and for each concept write: what it is (2–3 sentences) + its connection to the center (1 sentence) + supporting literature ID.
Inform the user that this session will perform about 8–10 retrievals, then start directly without waiting for confirmation.
Step 2 — Anchor the evolutionary lineage, perform tracing retrieval (lineage side) Using the same term as the topic, first find the current landmark work as the anchor, then trace backward layer by layer (3 layers by default), and finally supplement the latest progress from the past 1–2 years. Label each inheritance edge with its relationship type (method source / problem source / evidence source / field turning point).
Step 3 — Cross-identify hub nodes Compare the retrieval results from both sides and findconcepts that appear simultaneously in the conceptual map and the evolutionary lineage—these are hub nodes, the concepts that need to be mastered with the highest priority to understand this field. Hub nodes are usually both core neighbors on the map and key turning points in the lineage.
Step 4 — Comprehensive judgment Based on the retrieval results from both sides, write a 2–4 sentence introductory framework description: what the core problem of this field is, what the current technological landscape revolves around, and which hub node is best to start from when entering this field. Do not make value judgments about whether "this field is good or not".
Deliverables
After retrieval ends, create in the current directory the {topic-slug}/ subfolder and output four things:
- Terminal —directly print the comprehensive summary: hub node list + overarching questions + recommended entry path
report.md—stable YAML front matter + comprehensive judgment + hub node explanations + concept adjacency list (including Mermaid diagram) + inheritance chain + recommended reading order + retrieval transparencyreport.html—comprehensive report, see the HTML specification belowreferences.bib—merge all queries'.bib, and after deduplication keep only the literature cited by the report
HTML report specification
Accent color #0891b2(Teal, same color family as bear-map, reflecting the extension of the "cognitive map").
First screen: Comprehensive judgment panel
.verdict-title: What is the overarching question this field has been answering all along
.verdict-note: Introductory framework, 2–4 sentences—what the core question is, what the main trunk of the current landscape is, and where to enter most efficiently
.metric-strip: Number of neighboring concepts / number of lineage nodes / number of hub nodes / number of recommended introductory references
Hub node panel(below the first screen, the most important section)
List 2–4 hub nodes; each hub node card includes:
- Concept name + labels for "role in the concept landscape" and "position in the lineage"
- What it is (2–3 substantive sentences)
- Why it is a hub (1 sentence: it plays a key role in both the landscape and the lineage)
- DOI link to a representative reference recommended for reading
Detailed content area (two tabs)
tab "Concept Landscape": SVG concept star map (center node filled in teal, neighboring nodes with white background and teal borders, with relationship labels); below the star map, each neighbor has a complete explanatory card (what it is, 2–3 sentences + 1 sentence on its connection to the center). Hub nodes are distinguished in the star map with stroke-width: 3; stroke: var(--accent) bold borders, and explanatory cards add .card-primary class。
tab "Evolutionary Lineage": vertical lineage timeline, with each node highlighting turning points, and three summary cards at the end (overarching question / open edges / retrieval gaps). Hub nodes are marked with a star in the timeline.
Reference cards in both tabs include DOI links and collapsible Chinese academic abstracts.
Recommended reading order(independent section at the bottom): 3–5 papers, explaining why to read in this order and which cognitive gap each paper addresses.
Retrieval transparency(bottom): list all query terms, mode, number returned, number adopted, retrieval gaps.
Honest boundaries
Hub nodes are intersections based on the results of this retrieval, not an authoritative classification of the complete field. Retrieval coverage is limited, and there may be gaps in the lineage chain. The introductory framework is a distillation of the retrieval results, not a complete evaluation of the field.
Gotchas
- Hub nodes must not be forced to make up the numbers: If the retrieval results on both sides have no true intersecting concepts, honestly say "this retrieval found no clear hub node," and separately recommend the highest-weighted concept on each side as the starting point for entry
- The retrieval target for the concept landscape and the lineage must be the same term: Do not search a broad field on the landscape side and a specific paper on the lineage side
- The comprehensive judgment must be based on the retrieved content from both sides, not merely impressions from the concept landscape side