| name | explore |
| description | Use when the user wants to bootstrap a new topic, expand an existing topic boundary, or build a reading roadmap — "开一个新的 topic"、"我想系统读一下 X 领域"、"梳理这个领域的脉络"、"这是一批种子论文帮我建立 topic"、"expand the landscape for Y". Use even if the user does not say /explore. |
/explore
When To Use
- The user wants a new topic or a major topic expansion.
- The input is a deep-research list, a field overview, or a discussion about topic boundaries.
Inputs Expected
- topic name and slug
- user motivation and boundary hints
- optional: seed paper list
Preflight Checks
- New topic vs. expansion — decide before doing anything:
- New topic: no
content/topics/<slug>/ directory exists; or the new direction's papers have minimal overlap with all existing topics
- Expansion: the topic directory already exists with
index.md/overview.md; new papers clearly belong within the existing scope — only landscape.md and overview.md need updating, skip create_topic.py
- Confirm the canonical topic slug (use underscores, e.g.,
causal_representation_learning).
- Decide whether the user wants topic ordering updated in the Quartz explorer (
quartz.layout.ts).
Workflow
For a new topic
- Run
scripts/create_topic.py "Human-Readable Title" --slug <slug> --description "..." to scaffold index.md, overview.md, and landscape.md.
- Fill in
landscape.md using the structure below.
- Fill in the
Scope section of topic-index.md to articulate boundaries with neighboring topics.
- Batch the supplied seed papers through
/intake.
- Rebuild indexes.
- Append ops log entry (
event=explore).
For a topic expansion
- Read the existing
landscape.md and overview.md first.
- Update the relevant sections of
landscape.md (add new sub-areas, update roadmap).
- Batch new papers through
/intake.
- Rebuild indexes if needed.
- Append ops log entry.
Landscape.md Structure
landscape.md should address all five sections. Stubs are acceptable initially, but each section must have at least one substantive paragraph before the topic is considered bootstrapped:
- Problem Space: What fundamental problem(s) this field is trying to solve; why they are hard; what past approaches failed and why
- Methodology Spectrum: Current technical approaches from the most conservative to the most aggressive, laid out as a spectrum; each major approach labeled with 1–2 representative papers
- Evolution: A rough timeline with 3–5 key turning points; each annotated with the representative paper(s) that caused or marked the shift
- Key Open Questions: 3–5 specific, concrete unsolved problems recognized by the community — not vague aspirations like "improve generalization"
- Reading Roadmap: Recommended reading order for someone new to this topic, divided into
entry / core / advanced tiers
Neighboring Topic Boundaries
When a new topic overlaps with an existing one:
- Write explicit boundary statements in both topics'
index.md Scope sections
- Use cross-references (
[[topic-slug/index]]) to link related topics
- If overlap is very high (most papers belong in both), consider whether a merge or a cross-cutting synthesis page is more appropriate than a new topic
Files It May Write
content/topics/<topic>/index.md
content/topics/<topic>/overview.md
content/topics/<topic>/landscape.md
quartz.layout.ts (if topic ordering is updated)
ops/log.md
Scripts
assets/topic-index.md, assets/overview.md, assets/landscape.md — templates
scripts/create_topic.py — scaffolds all three files
../intake/scripts/intake.py — for seeding the paper list
../maintain/scripts/rebuild_index.py
- See
_shared/script-registry.md for full invocation examples
Done Criteria
- The topic folder is browsable with
index.md and overview.md
landscape.md has substantive content in all five sections
- The seed paper list has been queued via
/intake
- Topic boundary is documented in
index.md Scope section
Hand-Offs
- Cross-topic comparison work →
/synthesize
- Idea formation during exploration →
/discuss