agentleFS
Felpix-StudiosFelpix-Studios/social-science-research/skills/lit-review/SKILL.md

lit-review

Skill8 starsChanged 4 months agoMITView on GitHub

Structured literature review via parallel Librarian agents. Searches top journals, working paper repos (NBER, SSRN, IZA); traces citation chains from key papers. Use when surveying existing research on a topic, not finding datasets or writing a paper.

20 other projects use a version of this, 2 of them edited.

Adopted by

Most recently taken up by

Used with

Projects with it also use research-ideation, devils-advocate and 2 more.

Other files people use alongside this one.

Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
  • luischanciSkill

    research-ideation

    Generate structured research questions, testable hypotheses, and empirical strategies from a topic or dataset.

    Also in 4 projects that use this

  • bidodevSkill

    devils-advocate

    Challenges decisions, plans, and code by systematically arguing the opposing side. Surfaces risks, pokes holes in assumptions, and stress-tests thinking before committing. Supports intensity levels (gentle, balanced, ruthless, linus) and works with conversation context or a specific file reference. Trigger on requests like "challenge this", "poke holes", "what could go wrong", "play devil's advocate", "linus mode", or "/devils-advocate".

    Also in 4 projects that use this

  • seikaikyoSkill

    interview-me

    需求不明時的意圖萃取訪談:一次一題、每題附上自己的猜測、聽出「真正想要 vs 覺得應該要」,直到能預測使用者反應(約 95% 信心)才動工。適用:需求缺少對象 / 動機 / 成功標準 / 約束,或使用者點名「訪談我」「先確認一下」「我們確定嗎」。明確自足的指示、純資訊查詢、機械性操作不適用。

    Also in 4 projects that use this

  • SentrySkill

    commit

    Create commit messages following Sentry conventions. Use when committing code changes, writing commit messages, or formatting git history. Follows conventional commits with Sentry-specific issue references.

    Also in 4 projects that use this

The file

We split this skill into the rules it gives an agent

7 rules. 0 are common, 1 is distinctive, and 6 are very specific to this project.

  • 0

    Common

    Most of the 18 edited versions we read keep these.

  • 1

    Distinctive

    This author's own choices. Worth a look.

  • 6

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Read quality_reports/project_spec_*.md or quality_reports/research_*.md if they exist — extract the research question and identification …

    line 18
  2. Very specific:

    Check references/papers/ — list any PDFs already on hand.

    line 19
  3. Very specific:

    Read the project .bib file (Bibliography_base.bib or any .bib at project root) — extract paper titles and authors already known.

    line 20
  4. Very specific:

    Read references/domain-profile.md if it exists — get the field's journal list and key researchers.

    line 21
  5. Very specific:

    If references/domain-profile.md does not exist, use AskUserQuestion to determine the field:

    line 24
  6. Distinctive:

    After all Librarians complete:

    line 93
  7. Very specific:

    Recommend /validate-bib: after the user incorporates new citations into the manuscript.

    line 199

“Common” is measured: kept by at least 6 in 10 of the 18 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Literature Review

Conduct a systematic literature review using a parallel fleet of Librarian agents, each searching a different angle. Citation chains from anchor papers are the highest-yield search vector and always run when anchor papers are available.

Input: $ARGUMENTS — a topic, research question, or the title of a known key paper.

Step 1: Orient and Gather Context

  1. Read quality_reports/project_spec_*.md or quality_reports/research_*.md if they exist — extract the research question and identification strategy.
  2. Check references/papers/ — list any PDFs already on hand.
  3. Read the project .bib file (Bibliography_base.bib or any .bib at project root) — extract paper titles and authors already known.
  4. Read references/domain-profile.md if it exists — get the field's journal list and key researchers.
  5. Identify anchor papers: the 1–3 most central known papers (from existing bib or supporting docs). These seed the citation chain search.

If references/domain-profile.md does not exist, use AskUserQuestion to determine the field:

  • header: "Field"
  • question: "What field is this research in? This determines which journals and repositories the librarian agents search."
  • options:
    • label: "Economics", description: "AER, QJE, Econometrica, JPE, ReStud + NBER/IZA"
    • label: "Political Science", description: "APSR, AJPS, JOP, CPS, World Politics"
    • label: "Sociology", description: "ASR, AJS, Social Forces, Demography"
    • label: "Public Health", description: "NEJM, Lancet, JAMA, AJE"

Step 2: Assign Librarians

Dispatch 3–5 Librarian agents in parallel using Task. Each gets a specific search angle, the topic, and the anchor papers (if any).

Always dispatch (3 minimum):

Librarian 1 — Top Journals:

Task prompt: "You are a Librarian agent. Topic: [topic]. Search angle: Top Journals.
Field: [field from domain-profile or user answer].
Top journals to search: [list from domain-profile, or infer for field].
Anchor papers (for context only): [list].
Find the 8-12 most relevant published papers from the past 10 years.
For seminal papers, go back further. Follow the Librarian agent instructions."

Librarian 2 — Working Papers (NBER + SSRN + IZA):

Task prompt: "You are a Librarian agent. Topic: [topic]. Search angle: NBER + SSRN + IZA.
Search NBER (site:nber.org), SSRN (site:ssrn.com), and IZA (site:iza.org/publications/dp)
for recent working papers on this topic. Find 6-10 working papers.
Flag all as working papers. Follow the Librarian agent instructions."

Librarian 3 — Secondary Journals + Adjacent Fields:

Task prompt: "You are a Librarian agent. Topic: [topic]. Search angle: Secondary Journals.
Field: [field]. Secondary journals: [list from domain-profile, or search broadly].
Also search adjacent fields that might have relevant methodology or findings.
Find 5-8 papers. Follow the Librarian agent instructions."

Add when anchor papers are available (dispatch 4th and 5th):

Librarian 4 — Backward Citation Chain:

Task prompt: "You are a Librarian agent. Topic: [topic]. Search angle: Citation Chain — BACKWARD.
Anchor papers: [title, authors, year for each].
For each anchor paper: use Semantic Scholar API to find what it cites.
Extract 8-12 most relevant references from the combined reference lists.
These are the papers the anchor papers build on — the theoretical and empirical foundations.
Follow the Librarian agent instructions for citation chain search."

Librarian 5 — Forward Citation Chain:

Task prompt: "You are a Librarian agent. Topic: [topic]. Search angle: Citation Chain — FORWARD.
Anchor papers: [title, authors, year for each].
For each anchor paper: use Semantic Scholar API to find who cites it.
Focus on papers published in the last 3 years — these represent the active research frontier.
Find 8-12 most relevant papers that cite the anchor papers.
Follow the Librarian agent instructions for citation chain search."

Why citation chains beat keyword search: The papers that cite your anchor papers are by definition working on the same problem. The papers your anchor papers cite are the foundations your paper must engage with. These vectors find papers that keyword searches miss entirely.

Step 3: Consolidate and Deduplicate

After all Librarians complete:

  1. Merge all paper lists from all Librarians.
  2. Deduplicate: if the same paper appears from multiple sources, keep one entry and note "Found via: [multiple sources]" — cross-source confirmation increases confidence.
  3. Flag any papers marked FLAG by a Librarian — do not include in final BibTeX without manual verification.
  4. Total target: 20–40 papers (quality over quantity).

Step 4: Organize Thematically

Sort papers into these categories:

  • Theoretical Contributions — models, mechanisms, frameworks
  • Empirical Findings — key results, effect sizes, data used
  • Methodological Innovations — estimators, identification strategies, inference
  • Open Debates — unresolved disagreements, conflicting findings

For each category, write a 2–3 paragraph synthesis (not just a list).

Step 5: Identify Gaps

List 3–5 specific gaps the reviewed literature leaves open:

  • Questions not asked
  • Populations or contexts not studied
  • Data limitations that constrained prior work
  • Methodological assumptions no one has relaxed
  • Conflicting findings that need resolution

For each gap: "No paper has examined X in context Y using method Z. This matters because..."

Step 6: Save Report

Save to quality_reports/lit_review_[sanitized_topic].md:

# Literature Review: [Topic]

**Date:** [YYYY-MM-DD]
**Search strategy:** [N Librarians — journals, NBER/SSRN/IZA, citation chains from: [anchor papers]]
**Papers reviewed:** N total (M published, K working papers)

## Summary

[3-paragraph overview of the state of the literature]

## Key Papers

### [Author(s) (Year)] — [Short Title]
- **Venue:** [Journal / NBER wXXXX / IZA DP XXXX]
- **Method:** [Identification strategy + data]
- **Key finding:** [Result with effect size]
- **Relevance:** [Why it matters for our question]
- **Found via:** [Librarian source]

[Repeat for all papers, ordered by relevance]


## Thematic Organization

### Theoretical Contributions
[Synthesis paragraph(s)]

### Empirical Findings
[Synthesis comparing results across studies]

### Methodological Innovations
[Methods relevant to our research question]

### Open Debates
[Conflicting findings and unresolved questions]


## Gaps and Opportunities

1. [Gap 1 — specific and actionable]
2. [Gap 2]
3. [Gap 3]


## Suggested Next Steps

- Papers to read in full: [list 3-5 highest priority]
- `/data-finder [topic]` — identify datasets for the priority research question
- `/validate-bib` after incorporating new citations into the manuscript


## BibTeX Entries

```bibtex
[all verified entries]

Flagged Papers (Unverified — Manual Check Required)

[Papers Librarians marked FLAG — verify before citing]



## Important

- **Working papers change.** Note the date retrieved for all working papers.
- **Do not include flagged papers in the BibTeX block.** Move them to the Flagged section.
- **Citation chains are the most productive vector.** Librarians 4 and 5 often find the most important missing papers — always dispatch them when anchor papers are available.
- **Recommend `/validate-bib`** after the user incorporates new citations into the manuscript.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.