agentleFS
Jeng-linLiJeng-linLi/humanities-research-skills/skills/lit-review/SKILL.md

lit-review

Skill0 starsChanged 49 days agoMITView on GitHub

Systematic or narrative literature review with PRISMA flow, auditable screening decisions, and citation-traceable synthesis. Use for literature review, 文獻回顧, systematic review, scoping review, or when the user needs to search, screen and synthesise academic sources.

20 other projects use a version of this.

Adopted by

Most recently taken up by

Used with

Projects with it also use review-r, validate-bib and 7 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

    review-r

    Run the R code review protocol on R scripts. Checks code quality, reproducibility, domain correctness, and professional standards. Produces a report without editing files.

    Also in 4 projects that use this

  • luischanciSkill

    validate-bib

    Validate bibliography entries against citations in all manuscript files. Find missing entries and unused references.

    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

  • shanggqmSkill

    data-analysis

    数据分析技能,提供结构化的数据查询、统计分析和报告生成工作流。当用户需要查询数据库、进行数据统计分析、生成分析报告时使用。按规范输出原始数据、计算结果、Markdown报告和HTML可视化报告。

    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

  • +3 more from luischanci

The file

We split this skill into the rules it gives an agent

8 rules. 0 are common, 5 are distinctive, and 3 are very specific to this project.

  • 0

    Common

    Most of the 18 edited versions we read keep these.

  • 5

    Distinctive

    This author's own choices. Worth a look.

  • 3

    Very specific

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

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

    Before searching anything, write protocol.md:

    line 13
  2. Distinctive:

    Do not start searching until the user approves the protocol.

    line 20
  3. Distinctive:

    Keep findings as the authors stated them.

    line 52
  4. Very specific:

    every paragraph ends with claim ids, e.g.

    line 66
  5. Very specific:

    No claim without a source_id and a locator.

    line 82
  6. Distinctive:

    No fabricated counts. Flow numbers are computed from the logs, not estimated.

    line 83
  7. Distinctive:

    Never convert a paraphrase into a quotation.

    line 84
  8. Distinctive:

    If the user asks for a claim the corpus doesn't support: say it, and name the evidence that would be needed.

    line 85

“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

Produces a review where every claim is traceable and every exclusion has a reason. Read references/evidence-contract.md first — it is binding.

Step 0 — Freeze the protocol first

Before searching anything, write protocol.md:

  • research question (one sentence, PICO/PECO or topic + lens if not applicable)
  • inclusion / exclusion criteria, written so a stranger would apply them the same way
  • databases and date range
  • the search string, verbatim

Do not start searching until the user approves the protocol. Changing criteria after seeing results is the most common way reviews become unsound.

Step 1 — Search

Record in search-log.md, per database:

database, platform/interface, date run, exact query string, hits, exported

Save the raw export. The query string must be reproducible — a review that cannot be re-run cannot be published.

Step 2 — Screen

Two passes, both logged:

  1. title + abstract → include / exclude / uncertain
  2. full text → include / exclude with a stated reason

Output screening-log.csv: source_id, pass1, pass2, reason, decided_by.

Uncertain records are never silently dropped. Surface them to the user with the ambiguity named.

Step 3 — Extract

extraction-table.csv, one row per included source, columns agreed in step 0 (typically: id, authors, year, context, method, sample, key finding, limitations, locator).

Keep findings as the authors stated them. Do not pre-synthesise in this table.

Step 4 — Synthesise

Choose one and say which:

  • narrative / thematic — group by theme, argue across sources
  • tabular — structured comparison
  • framework — build or extend a conceptual structure

For qualitative synthesis, use qual-coding on the extracted findings.

Rules:

  • every paragraph ends with claim ids, e.g. [C012, C017]
  • report disagreements between sources, do not average them away
  • state what the corpus cannot answer

Step 5 — Report

Write review.md:

  1. question and protocol (abbreviated)
  2. PRISMA-style flow counts: identified → deduplicated → screened → full-text assessed → included, with exclusion reasons and counts
  3. synthesis
  4. limitations, including your own search limits
  5. evidence table (see contract §5)

Hard constraints

  • No claim without a source_id and a locator.
  • No fabricated counts. Flow numbers are computed from the logs, not estimated.
  • Never convert a paraphrase into a quotation.
  • If the user asks for a claim the corpus doesn't support: say it, and name the evidence that would be needed.

Files

protocol.md
search-log.md
screening-log.csv
extraction-table.csv
review.md
sources.yml

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.