
from scischolar38
Turn SciScholar search-papers output into a structured, evidence-grounded literature review, including an evidence ledger, thematic clusters, and reading path.
This skill orchestrates SciScholar's search-papers results into a full literature review deliverable. It enforces a safe operating contract (only use search-papers), builds an evidence ledger of papers (title, problem, method, evidence, role), clusters findings into thematic sections, and produces a recommended reading path and gap analysis. Outputs include the exact search command(s) (token omitted), an evidence table, and a thematic review.
Use this when a user asks for a literature review, related work section, survey outline, or a curated reading list on a research topic. It triggers on phrases like "literature review", "related work", "paper survey", "research reading list", or "topic overview".
Best for agents capable of running CLI tools and managing tokens securely (research assistants, SciScholar-enabled agents, LLM agents with shell access).
Instruction-only skill (no bundled scripts) for conducting literature reviews using the SciAtlas CLI search-papers command. The repo was renamed from scischolar to SciAtlas, and the SKILL.md URL in the DB is now 404. The actual SKILL.md is well-structured with clear bootstrap, search, and review construction steps. Security is good — instructs agents never to include API tokens in output, and all network calls go to the product's own API. Minor deduction for pip install from GitHub without checksum or version pinning.
Source URL (https://github.com/zjunlp/scischolar/blob/main/agent-skill/scischolar-literature-review/SKILL.md) is broken due to repo rename to SciAtlas. Actual SKILL.md found at https://github.com/zjunlp/SciAtlas/blob/main/agent-skill/sciatlas-literature-review/SKILL.md. The slug and URLs in the DB should be updated. Skill is prompt-only with no executable scripts.