brave-search
About
This skill provides structured Brave Search web queries and summarization workflows for MCP-style agents, enabling fact-finding on current events and research. It retrieves SERP data (web pages, FAQs, news) and can generate concise narrative summaries when authorized. Developers need to configure the `BRAVE_SEARCH_API_KEY` environment variable to use it.
Quick Install
Claude Code
Recommended/plugin add https://github.com/akaihola/skills-akaiholagit clone https://github.com/akaihola/skills-akaihola.git ~/.claude/skills/brave-searchCopy and paste this command in Claude Code to install this skill
Documentation
Purpose
Provide deterministic wrappers around Brave Search’s web and summarizer endpoints. Use the web workflow to retrieve structured SERP data (web pages, FAQs, discussions, news, videos). Use the summarizer workflow to turn Brave’s aggregated findings into a concise narrative when the subscription permits summarization.
When to Use
- Run web fact-finding on current events, product comparisons, research digests, or perspective gathering when Google/Bing responses are insufficient.
- Request the summarizer only after a prior web search produced a
summarizer_key, and the user explicitly wants a Brave-generated synthesis.
Configuration Requirements
- Set
BRAVE_SEARCH_API_KEYin the execution environment. The value populates theX-Subscription-Tokenheader. - Prefer secure storage through the project’s secrets tooling before launching the script.
- All invocations must use
uv runto respect the project’s Python environment.
Workflows
A. Web Search (results only)
- Prepare JSON containing at least
"query". Optional keys includecountry,search_lang,ui_lang,count,offset,safesearch,freshness,text_decorations,spellcheck,result_filter,goggles,units, andextra_snippets. - Run
uv run scripts/brave_search.py web --params-json '<JSON>'. - Consume
web_results,faq_results,discussions_results,news_results, andvideo_resultsfrom the JSON output. Each section mirrors the Brave MCP tool’s simplified records. - If
okisfalsewith"No web results found", broaden or restate the query before retrying.
B. Web Search with Summarizer Key
- Follow workflow A but add
"summary": trueto the JSON payload. - The script automatically requests
result_filter=summarizer. Inspect the response’ssummarizer_key. - Store the key and cite the original
web_resultswhen answering detailed questions while preparing for a summarizer follow-up.
C. Summarizer
- Ensure a recent workflow B run produced a
summarizer_key. - Build JSON like
{"key": "<summarizer_key>", "entity_info": false, "inline_references": true}. Optional overrides:poll_interval_ms(default 50) andmax_attempts(default 20). - Run
uv run scripts/brave_search.py summarizer --params-json '<JSON>'. - Use
summary_textas the main synthesis. Supplement withenrichments,followups, andentities_infosfor deeper context or suggested next steps. - If the summarizer fails, rely on the previously collected
web_resultsto craft a manual answer.
Error Handling and Fallbacks
- Missing API key: the script emits
ok: falsewith an explicit description; set the environment variable and rerun. - HTTP or Brave-side errors: review the
detailsobject, adjust parameters, or pause if throttled. - Summarizer polling timeout: rerun the web search to refresh the key, or answer using raw web data.
References
references/brave_web_search_params.md— exhaustive parameter definitions and sample payloads.references/brave_summarizer_workflow.md— polling logic, summary message schema, and tuning guidance.references/brave_search_examples.md— end-to-end scenarios demonstrating combined web and summarizer usage.
GitHub Repository
Related Skills
sglang
MetaSGLang is a high-performance LLM serving framework that specializes in fast, structured generation for JSON, regex, and agentic workflows using its RadixAttention prefix caching. It delivers significantly faster inference, especially for tasks with repeated prefixes, making it ideal for complex, structured outputs and multi-turn conversations. Choose SGLang over alternatives like vLLM when you need constrained decoding or are building applications with extensive prefix sharing.
Algorithmic Art Generation
MetaThis skill helps developers create algorithmic art using p5.js, focusing on generative art, computational aesthetics, and interactive visualizations. It automatically activates for topics like "generative art" or "p5.js visualization" and guides you through creating unique algorithms with features like seeded randomness, flow fields, and particle systems. Use it when you need to build reproducible, code-driven artistic patterns.
business-rule-documentation
MetaThis skill provides standardized templates for systematically documenting business logic and domain knowledge following Domain-Driven Design principles. It helps developers capture business rules, process flows, decision trees, and terminology glossaries to maintain consistency between requirements and implementation. Use it when documenting domain models, creating business rule repositories, or bridging communication between business and technical teams.
huggingface-accelerate
DevelopmentHuggingFace Accelerate provides the simplest API for adding distributed training to PyTorch scripts with just 4 lines of code. It offers a unified interface for multiple distributed training frameworks like DeepSpeed, FSDP, and DDP while handling automatic device placement and mixed precision. This makes it ideal for developers who want to quickly scale their PyTorch training across multiple GPUs or nodes without complex configuration.
