arpsych-literature-synthesis
GitHub用于系统性地搜集、阅读并结构化心理学年度综述文献，构建全面且可追溯的证据矩阵。通过种子文献、引文追踪及多数据库检索确保覆盖度，记录搜索协议以支持透明度，为后续分析框架奠定基础。
Trigger Scenarios
Install
npx skills add brycewang-stanford/Awesome-Journal-Skills --skill arpsych-literature-synthesis -g -y
SKILL.md
Frontmatter
{
"name": "arpsych-literature-synthesis",
"description": "Use when systematically gathering, reading, and structuring the literature for an Annual Review of Psychology (ARPsych) review so coverage is comprehensive and current. Builds the evidence base; it does not design the analytical spine (arpsych-organizing-framework) or judge balance (arpsych-comprehensiveness-and-balance)."
}
Literature Synthesis — Reading the Whole Field (arpsych-literature-synthesis)
When to trigger
- An invitation is in hand and the reading must start in earnest
- The user has "read a lot" but has no system for tracking what is covered
- Coverage gaps are likely (one paradigm, one lab, or one decade dominates the notes)
- An embedded meta-analysis is planned and the search must be documentable
Coverage is the price of admission, not the contribution
ARPsych readers trust a review to be comprehensive within its frame and current to the frontier. A reviewer caught missing a major line of work loses authority instantly. But coverage is the floor: the contribution is the framework you impose later (arpsych-organizing-framework). This skill builds an evidence base broad and well-recorded enough that (a) nothing important is missed and (b) you can later be selective in prose without being incomplete in coverage.
A systematic, documentable search
Even a narrative review benefits from a near-systematic search you can describe:
- Seed set — the canonical and recent agenda-setting papers you already know.
- Forward + backward citation chasing — who cites the seeds; whom the seeds cite.
- Database sweep — PsycINFO/APA PsycNet, Web of Science, Scopus, Google Scholar with explicit terms.
- Specialty + adjacent venues — the relevant primary journals AND adjacent fields (neuro, economics, linguistics) the topic touches.
- Stopping rule — saturation: new searches surface only papers already in the matrix.
- Record the protocol — terms, databases, dates, inclusion/exclusion — even for a narrative review; mandatory if you report a meta-analysis (→
arpsych-transparency-and-reproducibility).
Document the search as you go, not from memory afterward. ARPsych's transparency expectations and any meta-analytic component both depend on a search log you can reproduce (检索于 2026-06；以官网为准).
The evidence matrix (one row per study/finding)
Build a structured table — not a pile of PDFs. Useful columns:
| Column | Why |
|---|---|
| Construct / question | lets you later sort papers into framework cells |
| Population / paradigm | surfaces over-reliance on one method or sample (WEIRD bias) |
| Design + strength | feeds credibility weighting in arpsych-comprehensiveness-and-balance |
| Effect / finding (+ direction, size if available) | enables quantitative synthesis |
| Replication status | flags shaky vs. robust effects (post-replication-crisis hygiene) |
| Lab / camp | exposes citation concentration and balance risks |
The matrix is the bridge: its rows become framework cells, its columns become balance and credibility checks.
Checklist
- Seed set, citation chasing, database sweep, and adjacent venues all run
- Search terms, databases, and dates recorded (protocol, not memory)
- Saturation reached (new searches return only known papers)
- Evidence matrix built: one row per study/finding with the columns above
- Replication status noted for contested effects
- Lab/camp column reveals whether coverage concentrates on a few groups
- Non-English / non-WEIRD and cross-disciplinary work checked for, not assumed away
- Coverage is current to the frontier AND anchored to foundational work
Anti-patterns
- Reading by familiarity (only your own subfield's journals) — guarantees coverage gaps
- No search protocol — cannot satisfy ARPsych transparency or support a meta-analysis
- A pile of summaries instead of a structured matrix — no path to a framework
- Ignoring replication status, so debunked effects are reviewed as established
- Stopping when tired rather than at saturation
- Treating coverage as the contribution (it is the floor, not the spine)
Output format
【Search protocol】seeds + databases + terms + dates + stopping rule (recorded?)
【Coverage】subfields/paradigms/adjacent fields swept; saturation reached? Y/N
【Evidence matrix】#rows; columns include design, replication status, lab/camp? Y/N
【Gaps / concentration】where coverage thins or over-concentrates
【Meta-analysis planned?】Y/N → if Y, route to transparency skill
【Next step】→ arpsych-organizing-framework (turn the matrix into a spine)
Version History
- 1839142 Current 2026-07-05 12:24
