rune-research

Executes multi-source web searches and synthesizes cited findings with confidence levels.

1|Updated Mar 22, 2026
One-click install
npx skills add https://github.com/dangvu008/VietTruyen --skill rune-research-dangvu008
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: rune-research
Source: https://github.com/dangvu008/VietTruyen/tree/main/.agents/skills/rune-research
Command: npx skills add https://github.com/dangvu008/VietTruyen --skill rune-research-dangvu008

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Getting reliable answers from the web requires more than a single search: single-source answers are often outdated, biased, or wrong. This Skill runs structured web research with source triangulation so every conclusion is backed by multiple independent sources and clearly labeled with a confidence level. ## Core Features & Use Cases - Multi-Query Search: Generates 2-3 varied search queries per question and enforces a minimum of 3 complementary sources from different source types (official docs, community, blogs, repositories). - Triangulated Synthesis: Cross-checks findings across sources, explicitly flags conflicting information, and assigns high/medium/low/unverified confidence ratings. - Saturation Detection: Stops searching when new queries yield under 10% new entities, saving tool calls and context tokens. - Use Case: Ask it to research "best practices for rate limiting in FastAPI" and receive a structured report with cited findings, conflicts between sources, code examples, and actionable recommendations. ## Quick Start Ask the agent to research a technical question such as "research the best approach for handling WebSocket reconnection in production" and receive a cited findings report.

Frequently Asked Questions about rune-research

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I research a technical question with cited sources?

Provide a research question and an optional focus area such as security or performance. The skill generates 2-3 search queries, fetches the top 3-5 URLs across diverse source types, and returns findings with a source URL attached to every claim.

How does web research avoid single-source misinformation?

It enforces a hard gate requiring at least 3 complementary sources from different source types (official docs, community, blogs, repositories). Single-source conclusions are automatically flagged as low confidence, and conflicting claims are reported explicitly.

What do the confidence levels in research results mean?

High means 3+ sources from different types agree, medium means 2 sources agree or 3+ from the same type, low means a single source or unresolved conflict, and unverified means no sources were found. The skill never fabricates findings when sources are missing.

Is there a limit on how many pages get fetched per research request?

Yes, a hard limit of 5 WebFetch calls per invocation. The skill also detects diminishing returns and stops early when new searches add under 10% new entities or results overlap heavily with prior searches.

When should I use web research instead of answering from model knowledge?

Use it when verifying package existence on npm or PyPI, checking current best practices, comparing competitor solutions, or any case where recency matters. Model knowledge can be outdated, while this skill cites 2025-2026 sources with URLs.