search-wolfram

Wolfram Resource Search

Systematically search the Wolfram ecosystem for resources related to a topic. This is a standalone skill — invoke it when the user wants to find what Wolfram already offers for their subject. Not every project needs this; it's for projects that build on or connect to the Wolfram Language ecosystem.

What you need

1. Topic / keywords — what to search for. Can be a research topic ("discrete Ricci curvature"), a function area ("graph distances"), or a concept ("hypergraph rewriting").
2. Project directory (optional) — where to put downloaded resources. If Wiki/ exists, resources are registered via add-resource. If not, results are just reported.

Kernel execution (license-aware)

The search scripts below each spawn a fresh wolframscript kernel, which consumes a license seat. Before running them, check headroom on the AgentTools MCP (no extra seat):

With[{free = $MaxLicenseProcesses - $LicenseProcesses}, free]

If free <= 0, do not spawn wolframscript — instead run the equivalent Import[...]/ResourceSearch[...] query through mcp__Wolfram__WolframLanguageEvaluator (and mcp__Wolfram__WolframLanguageContext for documentation), which reuse the one persistent kernel. The .wls scripts remain the path when the MCP is unavailable.

Sources

Run these scripts in parallel where possible. Skip any source that is clearly irrelevant to the topic.

1. Wolfram Language Documentation

Find built-in functions, guides, and tutorials relevant to the topic.

Script:

wolframscript -f "${CLAUDE_PLUGIN_ROOT}/scripts/search_wolfram_docs.wls" <keyword1> <keyword2> ...

Returns JSON array of {name, url, type} where type is "symbol", "guide", or "tutorial". All keywords must match (AND search) for symbol names. Guide and tutorial results come from a web search of the documentation site.

Also try: mcp__Wolfram__WolframLanguageContext with topic keywords for documentation context that includes usage patterns.

Output: Report the function list to the user. If Wiki/ exists, create a wiki article Wiki/Resources/WolframDocumentation_<Topic>.md summarizing the relevant functions and linking to their documentation pages.

2. Wolfram Function Repository

Search for community-contributed functions.

Script:

wolframscript -f "${CLAUDE_PLUGIN_ROOT}/scripts/search_function_repo.wls" <keyword1> <keyword2> ...

Returns JSON array of {name, url, description}. Uses ResourceSearch internally, filtered to Function type. Falls back to web scraping if ResourceSearch is unavailable.

Output: Report relevant functions. For each useful one, note the ResourceFunction["Name"] call. Create wiki article if Wiki/ exists.

3. Wolfram Community

Search the community forum for posts, discussions, and shared notebooks.

Script:

wolframscript -f "${CLAUDE_PLUGIN_ROOT}/scripts/search_wolfram_community.wls" <keyword1> <keyword2> ...

Returns JSON with {search_url, tag_url}. The Community site is JavaScript-rendered, so the script cannot scrape results directly. Use WebFetch on the returned URLs to get the actual search results.

What to extract from WebFetch results:

• Post titles and URLs
• Attached .nb files (download to Resources/)
• Code snippets in post bodies

Output: For each relevant post:

• If it has a .nb attachment: download to Resources/, create wiki article via add-resource
• If no attachment: create wiki article with URL and summary
• Title-match posts to topic keywords; skip loosely related ones

4. Stephen Wolfram's Writings

Search blog posts and essays for relevant material.

Script:

wolframscript -f "${CLAUDE_PLUGIN_ROOT}/scripts/search_wolfram_writings.wls" <keyword1> <keyword2> ...

Returns JSON array of {title, url, date}. Titles are extracted from URL slugs (lowercase, spaces).

Output: For highly relevant posts, create wiki articles with URL, summary, and key takeaways. Skip tangentially related posts — Wolfram writes about everything, so be selective.

5. Wolfram Physics Project

Only relevant if the project relates to Wolfram models, hypergraph rewriting, multiway systems, causal graphs, or related concepts. Skip otherwise.

Script:

wolframscript -f "${CLAUDE_PLUGIN_ROOT}/scripts/search_wolfram_physics.wls" <keyword1> <keyword2> ...

Returns JSON with:

• glossary — matching glossary terms from wolframphysics.org/glossary/
• archives — matching bulletins and working materials
• technical_documents — matching technical papers (arXiv PDFs, Wolfram Cloud docs)
• tools — available software tools
• urls — direct page URLs for all subsections (glossary, archives, tools, questions, universes, visual gallery, technical introduction)

Use WebFetch on the urls for deeper exploration of specific subsections (e.g., the full Technical Introduction, Q&A, universe registry).

After searching

Present a summary to the user:

Wolfram Resource Search: <topic>

Documentation:
- N relevant built-in functions found (list top ones)

Function Repository:
- N relevant community functions found

Community:
- N relevant posts found, M notebooks downloaded

Writings:
- N relevant blog posts found

Technical Introduction:
- Relevant / Not relevant (sections if relevant)

Suggest which resources are most useful for the project and what to explore next.

Integration with other skills

• Use add-resource for each downloaded notebook or notable reference
• Update Wiki/Index.md if new articles were created
• If Paper/ exists and references were found, suggest adding citations to Paper/references.bib