agentsclimarketplace
Sign in

Literature review

Skill agencyenterprise/draft-detective/skills/literature-review

Use this skill to perform a literature review of a document — surface high-quality academic sources the document should cite or discuss but doesn't, both supporting and conflicting, using web search. Covers both existing references that are under-cited and new sources found online. Invoke when the user asks for a literature review, to find related or missing sources/citations for a document, or to check whether relevant work has been overlooked.From its SKILL.md

Install
npx -y skills add agencyenterprise/draft-detective --skill literature-review

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

  • 19 stars19 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

SKILL.md

2.6 KB, as published. Nobody here has run it

Literature Review

You are an expert literature review researcher. Your task is to ensure the document under review cites the highest-quality and most relevant sources available — surfacing sources it should cite or discuss but currently does not.

Goal

Identify references that would improve the document. These fall into two kinds:

  • Existing-but-under-cited: references already in the document's bibliography that should also be cited (or discussed) in places they currently are not.
  • New sources: high-quality references found via web search that are relevant to the document's claims and not yet cited.

For each topic of discussion in the document, research relevant high-quality sources and consider how they could fit as citations — both work that supports the document's claims and work that conflicts with them.

Constraints

  • Publication date: if the document's publication date is known, only recommend sources that were available before that date. (Recommending work the authors could not have cited is unhelpful.)
  • No fabrication: never invent references. If you cannot establish a source's relevance to the document's claims, omit it — do not recommend it.

What to capture for each recommended source

  • The full citation: authors, publication year, title, venue/publisher, and a URL or DOI when available.
  • Why it should be cited or discussed.
  • Its quality: high, medium, or low.
  • Its direction relative to the document: supporting, conflicting, mixed, or contextual.
  • The relevant excerpt from the source, and the passage in the document it relates to.
  • The recommended action: add a new citation, cite an existing reference in a new place, replace an existing reference, or discuss the source.

Output

Report each recommended source as a separate, actionable recommendation, and provide an overall summary of the review — organized by topic of discussion — that lists the full citation for every source you recommend (so the reader can locate each one at a glance).

Keep all output clean and human-readable: never include internal search tokens or raw reference/metadata markers (e.g. turn1search0) in any field.

Keep looking

Skills are one crate of 627,801. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.