Agent Skills: web-research

Two-tier web research using web fetching as primary for known URLs and headless browser automation as fallback for searches. Requires 3+ sources, prioritizes official docs, includes versions/dates.

UncategorizedID: frostaura/ai.toolkit.gaia/web-research

Install this agent skill to your local

pnpm dlx add-skill https://github.com/frostaura/ai.toolkit.gaia/tree/HEAD/.github/skills/web-research

Skill Files

Browse the full folder contents for web-research.

Download Skill

Loading file tree…

.github/skills/web-research/SKILL.md

Skill Metadata

Name
web-research
Description
Two-tier web research using web fetching as primary for known URLs and headless browser automation as fallback for searches. Requires 3+ sources, prioritizes official docs, includes versions/dates.
<skill> <name>web-research</name> <description> A skill for conducting thorough web research using a two-tier approach: web fetching as the primary tool for known URLs, and headless browser automation as a fallback for searches and dynamic content. This skill ensures research quality through multiple sources, prioritizes official documentation, and includes versioning and publication dates. </description> <research-approach> <tier name="Primary" tool="Web Fetching"> <use-case>Known URLs (official docs, blogs, GitHub, API references)</use-case> <description>Fast and efficient for direct URL access without JavaScript rendering requirements.</description> <example>Use web fetching with URLs array and query string for known documentation URLs</example> <strengths> <strength>Fast execution</strength> <strength>Efficient for static content</strength> <strength>Ideal for known documentation URLs</strength> </strengths> <limitations> <limitation>No JavaScript rendering</limitation> <limitation>Cannot perform searches</limitation> </limitations> </tier> <tier name="Fallback" tool="Browser Automation"> <use-case>Search queries and dynamic content requiring JavaScript rendering</use-case> <description>Headless browser automation for complex interactions and search engine queries.</description> <example>Use browser automation tools for navigation, element interaction, keyboard input, and content extraction in interactive searches</example> <mode>Headless (no visible browser window)</mode> <strengths> <strength>Full JavaScript support</strength> <strength>Search engine integration</strength> <strength>Dynamic content handling</strength> </strengths> </tier> </research-approach> <quality-standards> <standard name="Sources"> <requirement>Minimum 3 sources for all claims</requirement> <priority>Official documentation preferred over third-party sources</priority> </standard> <standard name="Versioning"> <requirement>Include version numbers for all technologies</requirement> <format>v[X.Y.Z] or [Major].[Minor]</format> </standard> <standard name="Dates"> <requirement>Include publication or last-updated dates</requirement> <format>ISO 8601 or readable format (e.g., "January 2026")</format> </standard> <standard name="Caching"> <requirement>Cache findings using memory storage for future reference</requirement> <scope>ProjectWide for architectural decisions</scope> </standard> </quality-standards> <output-format> <template>Research report includes topic, recommendation with version and date, bulleted key findings, and full source URLs</template> <requirements> <requirement>Clear topic identification</requirement> <requirement>Explicit recommendation with version</requirement> <requirement>Bulleted key findings</requirement> <requirement>Full source URLs</requirement> </requirements> </output-format> <memory-integration> <workflow> <step name="Before Research"> <action>Recall topic from memory</action> <purpose>Retrieve existing knowledge to avoid redundant research</purpose> </step> <step name="After Research"> <action>Store research findings with sources as ProjectWide memory</action> <purpose>Cache findings for future reference and consistency</purpose> </step> </workflow> </memory-integration> <rules> <required> <rule>Try web fetching first for known URLs</rule> <rule>Cite all sources with full URLs</rule> <rule>Cache key findings via memory integration</rule> </required> <forbidden> <rule>Do not skip to browser automation for known URLs</rule> <rule>Do not present unverified information</rule> <rule>Do not cite fewer than 3 sources</rule> </forbidden> </rules> </skill>