Research search

Skill Axect/magi-researchers/skills/research-search

Three AI models, one synthesis — Claude, Gemini & Codex cross-verify each other for rigorous multi-perspective researchFrom the repository description

Install

npx -y skills add Axect/magi-researchers --skill research-search

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

13 stars13 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

2.9 KB, as published. Nobody here has run it

Research Search Skill

Description

Searches academic literature via OpenAlex (240M+ works) and optionally web sources. Standalone skill for quick literature discovery without the full brainstorm pipeline.

Usage

/research-search "topic" [--filter "..."] [--sort cited_by_count:desc|relevance_score:desc|publication_year:desc] [--limit N] [--web] [--save path/to/output.md]

Arguments

$ARGUMENTS — The search topic (required) and optional flags:
- --filter — OpenAlex filter string (default: "publication_year:>2021,type:article|review"). Valid types: article, review, preprint, book-chapter, book, dataset, dissertation, report.
- --sort — Sort order (default: relevance_score:desc). Options: relevance_score:desc, cited_by_count:desc, publication_year:desc.
- --limit — Number of results (default: 10, max: 50).
- --web — Also run a WebSearch for recent developments (default: false).
- --save — Save results to a file instead of displaying inline. Accepts a relative or absolute path.

Instructions

Shared rules: Read ${CLAUDE_PLUGIN_ROOT}/shared/rules.md before starting. §LaTeX applies to this skill.

When this skill is invoked, follow these steps:

Step 1: Parse Arguments

Extract the search topic from $ARGUMENTS.
Parse optional flags with defaults:
- filter: "publication_year:>2021,type:article|review"
- sort: relevance_score:desc
- limit: 10
- web: false
- save: null (display inline)

Step 2: Academic Literature Search (OpenAlex)

Run the bundled OpenAlex search script:

uv run python ${CLAUDE_PLUGIN_ROOT}/skills/research-brainstorm/scripts/openalex_search.py "{topic}" --filter "{filter}" --sort "{sort}" --limit {limit} --format md

If results are sparse (< 3 papers):

Simplify the query: drop domain-specific jargon, keep core concepts.
Retry with broader filter: --filter "publication_year:>2019,type:article|review|preprint".
If still sparse, retry without type filter: --filter "publication_year:>2019".

Step 3: Web Search (optional)

If --web is specified:

WebSearch(query: "{topic} recent advances 2024 2025 2026")

Extract the top 5 web results with source URLs.

Step 4: Format and Present Results

Combine results into a structured output:

## Literature Search: "{topic}"

### Academic Papers (OpenAlex)
**Filter**: {filter} | **Sort**: {sort} | **Results**: {count}

[numbered list from openalex_search.py output]

### Recent Developments (Web)
[key findings with source URLs — only if --web]

### Summary
- Total academic papers found: {N}
- Most cited: "{title}" ({year}, {citations} citations)
- Year range: {min_year}–{max_year}
- Dominant themes: [2-3 key themes from abstracts]

Step 5: Save or Display

If --save is specified: write the formatted output to the given path.
Otherwise: display inline in the conversation.