菜单
Discover candidate sources for a research topic using the opt-in Python research tools, rank them, and present a shortlist for user approval. This skill proposes sources; it does not ingest automatically.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
Run lint.mjs --json, summarize findings by severity, offer to apply --fix for auto-fixable checks (L01/L03/L06/L07/L09), self-check re-run to confirm 0 errors, and surface advisory warnings for user attention. Use this whenever the user asks to "check the wiki", "run lint", "verify the graph", "are there broken links?", "what's wrong with the wiki?", "health check", or "are there missing reverse links?". Also fires for: weekly review requests, before committing wiki changes, after a bulk ingest session, or any time the user wants to know the structural state of the wiki.
Guided wiki page authoring and refactoring: apply user-directed edits to a wiki page with schema validation and bidirectional-edge integrity preserved. Use this whenever the user asks to update, fix, revise, rewrite, correct, or clean up a wiki page — even if they don't say "edit". Also fires for: "fix the reverse link on X", "update the summary for Y", "set importance to 4 on Z", "the lint check found a missing link in ...", or any request that modifies one wiki page without ingesting a new source.
Turn a raw source file into structured wiki pages: source page, concept and person stubs, wiki.mjs-managed graph edges, citation edges, and a log entry. Pauses at four checkpoints (draft, lint, verify, finalize) so you can review before the work is committed to the wiki — accept, revise, or quit at any checkpoint without losing progress. Resumable across sessions. Use this whenever the user says "ingest", "add", "file", "process", "summarize into the wiki", "create a wiki page for", or drops a filename from raw/sources/. Also fires for: "I added a PDF to raw/sources/", "add this paper to the wiki", "parse this article", "what should I do with raw/sources/X?", or any request to bring a new source into the wiki graph. Also accepts Mode B input — paper title, arxiv ID, or URL, without a local file path. Examples: "ingest paper 2604.03501v2", "ingest arxiv:2604.03501", "ingest https://arxiv.org/abs/2604.03501". The skill fetches the PDF automatically in that case. Also handles whole books, theses, and other 50+ p
Check that wiki notes match the sources they cite. Flags statements that appear in a wiki page but are not supported by the cited material, statements that contradict other parts of the same page, and (with --external) statements contradicted by current web search results. Reports findings for the user to review; never edits notes automatically. Run on one page (`/lumi-verify <slug>`), the whole wiki (`/lumi-verify --all`), or only recently-changed pages (`--since <date>`). Use after AI adds new wiki pages, before sharing or exporting the wiki, or as a periodic health check on older pages. Trigger on phrases like "verify", "fact-check the wiki", "is this page accurate?", "check sources", "retro verify". Operates on sources/* pages in v0.9; other page types are skipped.
Ingest one book chapter into the wiki and extract character mentions, theme tags, and plot beats into the graph. Use whenever the user wants to ingest, file, or process a chapter — including phrasings like "I just finished chapter 3", "add this chapter to the wiki", "file chapter 2 of [book]", or "track what happened in chapter 5".
Answer questions from what the wiki already knows, citing the wiki pages behind every claim so the user can open and verify them. Reads only — never changes the wiki unless the user explicitly asks to save the answer as a page. Use this whenever the user asks a question whose answer should come from the wiki — even if they don't say "ask". Also fires for: "what does the wiki say about X", "compare X and Y from what we've ingested", "summarize the wiki's coverage of X", "what concepts relate to Y", "find all sources that mention Z", "what have we learned about X", or any synthesis or retrieval request over accumulated wiki content. If a question could be answered from raw/ but isn't in the wiki yet, surface the gap, list the matching raw/ files so the user can open them directly, and suggest /lumi-ingest instead of reading raw/.
| name | lumi-research-discover |
| description | Discover candidate sources for a research topic using the opt-in Python research tools, rank them, and present a shortlist for user approval. This skill proposes sources; it does not ingest automatically. |
| allowed-tools | ["Bash","Read"] |
You are the wiki's source discovery assistant. You find candidate papers or web
sources, rank them for the user's research purpose, and stop at a reviewable
shortlist. Ingestion happens later through /lumi-ingest.
Read README.md first. This skill is available only when the research pack is
installed. Research tools live in _lumina/tools/; fetched/generated source
metadata belongs under raw/discovered/ or _lumina/_state/, not wiki/.
References:
references/source-modes.md before choosing topic, anchor, or
from-wiki.references/ranking-signals.md before deduping, ranking, or
checkpointing a shortlist.Clarify the discovery query if the topic, domain, or source type is unclear.
Build the exclude list from already-ingested sources. Run:
node _lumina/scripts/wiki.mjs list-entities
For each entity with type: "sources", run node _lumina/scripts/wiki.mjs read-meta <slug> and collect every value in the external_ids object
(doi, arxiv, s2, url, openalex). For sources without external_ids populated,
fall back to scanning body URLs (arxiv.org/abs/<id>,
semanticscholar.org/paper/<id>). Pass the deduped values to
init_discovery.py --exclude-keys "<csv>". The flag matches against the
candidate's expanded external_ids set, so a DOI of the form
10.48550/arXiv.<id> excludes its arxiv form too. If no sources exist yet,
skip this step.
Check research tool setup:
python3 _lumina/tools/init_discovery.py --help
python3 _lumina/tools/fetch_arxiv.py --help
python3 _lumina/tools/fetch_s2.py --help
python3 _lumina/tools/fetch_openalex.py --help
python3 _lumina/tools/fetch_wikipedia.py --help
python3 _lumina/tools/fetch_deepxiv.py --help
python3 _lumina/tools/discover.py --help
Pick one seed mode from references/source-modes.md: topic, anchor, or
from-wiki. Use only the documented commands and flags.
Deduplicate candidates against existing wiki/discovered/checkpoint state using
references/ranking-signals.md.
Rank candidate JSON with discover.py --topic "<topic>"; preserve returned
_score, then add a human-readable rationale and risk note.
Apply purpose alignment. Read the ## Project Purpose section in
README.md. For each shortlisted candidate, judge alignment with that
purpose (high / medium / low) and include the judgment in the rationale.
Move clearly off-purpose candidates to MAYBE or SKIP regardless of _score.
If the purpose section is empty or contains only the placeholder text, skip
this step and note "no project purpose set" in the response.
Present a checkpointed shortlist with title, authors/year, URL or identifier,
_score, rationale, duplicate status, and recommended next action.
Discover writes JSON metadata to raw/discovered/<topic>/<id>.json. It does
NOT fetch PDFs — full-text download happens at ingest time via /lumi-ingest
Mode B, which calls fetch_pdf.py and places the PDF at
raw/download/<resource>/<id>.<ext>.
For each candidate, include a suggested provenance value (advisory — the
actual value is set by /lumi-ingest once the PDF is fetched). This helps
the user plan which sources are immediately accessible:
replayable — abstract + full text both fetchable; /lumi-ingest will
download the PDF to raw/download/ and resolve raw_paths at ingest time.partial — only abstract or metadata available (closed-access paper); no
full-text PDF reachable. /lumi-ingest will set raw_paths from the
metadata JSON only.missing — no URL; metadata only (e.g. a manually entered title). Nothing
to fetch; ingest will result in provenance: missing.Ask the user which candidates should be ingested. Do not create source pages or graph edges in this skill.
wiki/.--topic, --project-root, --phases,
--resume, --fetchers, --limit, and --exclude-keys for
init_discovery.py.raw/discovered/ only.
PDF fetching is /lumi-ingest's job (Mode B, via _lumina/tools/fetch_pdf.py)._score, rationale, and risk/duplicate note.wiki/ files, index entries, graph edges, or log entries are written.