mvanhorn/clawdbot-skill-parallel

Overview

This skill provides high-accuracy web search and research via the Parallel.ai API, optimized for AI agents that need rich excerpts and citations. It returns structured results with source URLs, titles, relevant excerpts, and publishing metadata to support evidence-based outputs. The skill supports multiple search modes to balance accuracy, latency, and cost for different research needs.

How this skill works

The skill calls the Parallel Python SDK to run searches and returns a structured response containing a search_id, an ordered list of results, and usage stats. Each result includes the page URL, title, publish date (when available), and one or more relevant excerpts that can be used as citation anchors. You can choose modes: one-shot for balanced accuracy, fast for lower latency/cost, or agentic for multi-hop complex research.

When to use it

• Deep research that requires cross-referenced facts and reliable citations
• Company, person, or competitor research where source snippets are needed
• Fact-checking or evidence-based reporting that demands direct excerpts
• Complex multi-hop queries that require synthesis across multiple pages
• Situations where you need higher accuracy than general-purpose search APIs

Best practices

• Start with mode="one-shot" for most queries; switch to "fast" for quick lookups and to "agentic" for complex multi-step research
• Request a moderate max_results (5–10) to get focused, high-quality excerpts without excess noise
• Use the excerpts[] field as citation text and include the url/title when presenting facts
• Check publish_date to assess source recency and prioritize newer evidence for time-sensitive topics
• Batch similar queries to reduce API overhead and track usage via the provided usage stats

Example use cases

• Produce a referenced summary comparing transformer architectures using agentic mode for multi-hop research
• Build a company profile with cited facts and publish dates for investor briefings
• Run fact-checks on social claims, extracting direct excerpts to support verdicts
• Fetch latest AI industry news in fast mode for a low-latency news feed
• Generate annotated bibliographies or research dossiers with linked sources