Researcher-profile-driven paper intake and literature workbench for academic workflows. Use this whenever the user wants to skim, deep-read, card, compare, synthesize, map research gaps, or build a literature review from papers, arXiv/AlphaXiv links, DOIs, PDFs, landing pages, or existing paper JSON / workbench artifacts. Normalize sources into `paper-record`, then route into scan, deep-read, card, synthesis, review, or compatibility modes (`json`, `interpret`, `xray`). Trigger even when the user only says things like “精读这篇”, “整合这几篇”, “找研究空白”, or “搭综述框架”.
Unified entrypoint for paper intake, strategic reading, multi-paper synthesis, and review construction.
Keep paper-record as the normalization layer. Do not merge high-level
analysis back into the normalized record.
Use this skill when the job is to:
Do not use this skill when the primary job is to implement a paper. In that
case, route to paper2code.
paper-record — normalized single-paper factsresearcher-profile — user research anchorpaper-deep-read — single-paper strategic analysis artifactliterature-synthesis — cross-paper integration artifactreview-outline — literature-review planning artifactdoi.org/... URLspaper-record JSONresearcher-profile, paper-deep-read, literature-synthesis, or
review-outline JSON$ARGUMENTS, the latest user message, or a
pasted JSON artifact.scripts/normalize_paper.py first.researcher-profile or collect only the missing fields.scan
deep-read
card
interpret
xray
json
paper-recordsynthesis
review
jsonscansynthesissynthesis and mark any gap mapping as provisionalFor any paper-like input, run:
python "$SKILL_DIR/scripts/normalize_paper.py" \
--source "<paper-source>" \
--lang "<lang>" \
--fulltext "<auto|prefer|never>"
Use --save only when the user asked to persist the normalized JSON.
Before deep-read, card, synthesis, or review, prefer a
researcher-profile.
If missing, collect only these fields:
research_fieldcore_questionthesis (optional)target_tierstageIf the user clearly wants no back-and-forth, proceed with a generic profile-light analysis and explicitly mark that personalization is limited.
If the user wants persistence, create or update the profile with:
python "$SKILL_DIR/scripts/workbench_io.py" init-profile \
--path "<profile-path>" \
--research-field "<field>" \
--core-question "<question>" \
--thesis "<optional-thesis>" \
--target-tier "<target-tier>" \
--stage "<stage>"
When the user asks to save a deep read, synthesis, or review plan, write a JSON artifact plus an optional Markdown or Org sidecar:
python "$SKILL_DIR/scripts/workbench_io.py" save-artifact \
--workspace "<workspace-dir>" \
--artifact-type "<paper-deep-read|literature-synthesis|review-outline>" \
--title "<artifact-title>" \
--payload-file "<json-payload-file>" \
--profile-path "<optional-profile-path>" \
--source-record "<path-to-paper-record>" \
--sidecar-file "<optional-md-or-org>"
作者观点 from 系统分析[信息待核实]synthesis and review must integrate arguments across papers rather than
serially summarizing each paperreview paragraphs must use PEEL as a micro-argument structure, not a
citation listdeep-read:
references/routing.md — source classification and routing logicreferences/schema.md — canonical paper-record contractreferences/artifacts.md — researcher-profile and higher-level artifactsreferences/migration.md — compatibility and alias mappingreferences/modes/json.md — machine-readable output rulesreferences/modes/interpret.md — lightweight explanation pathreferences/modes/xray.md — compact critique pathreferences/modes/scan.md — single-paper quick triagereferences/modes/deep-read.md — full single-paper deconstructionreferences/modes/card.md — literature card onlyreferences/modes/synthesis.md — cross-paper integrationreferences/modes/review.md — literature-review planning and writing