research-trigger

grandinh's avatarfrom grandinh

Natural language wrapper for research command - automatically triggers /research when users ask questions requiring deep research, best practices lookup, or multi-source information gathering

⭐1stars🔀0forks📁View on GitHub🕐Updated Nov 17, 2025
[Web Research]

When & Why to Use This Skill

The research-trigger skill acts as an intelligent natural language bridge to advanced research capabilities. It automatically detects when a user requires deep-dive information, best practices, or multi-source synthesis and triggers the /research command. This eliminates the need for manual command syntax, providing users with comprehensive, cited reports on complex technical and general topics through seamless intent recognition.

Use Cases

  • Technical Implementation Guides: Automatically initiate research for complex coding tasks, such as 'How do I implement JWT authentication?' to get step-by-step instructions.
  • Best Practices Lookup: Quickly retrieve industry-standard patterns for software development, such as 'What are the best practices for API versioning and caching?'
  • Technology Comparison: Evaluate competing tools or frameworks by asking questions like 'What are the pros and cons of different state management libraries?'
  • Documentation & Example Discovery: Find specific integration examples or documentation for third-party services, such as 'Search for examples of Sentry integration in Node.js.'
  • Deep-Dive Investigations: Conduct thorough explorations into specific topics or bugs by requesting the agent to 'Investigate performance optimization techniques for React apps.'
nameresearch-trigger
descriptionNatural language wrapper for research command - automatically triggers /research when users ask questions requiring deep research, best practices lookup, or multi-source information gathering
schema_version1.0

research-trigger

Type: ANALYSIS-ONLY DAIC Modes: DISCUSS, ALIGN, IMPLEMENT, CHECK (all modes) Priority: High

Trigger Reference

This skill activates on:

  • Keywords: "research", "find information", "look up", "investigate", "best practices", "how do I", "what are the options", "search for", "find examples", "discover", "explore"
  • Intent Patterns: (research|investigate|find|lookup).*?, how.*(do I|implement|use|approach), what.*(options|alternatives|ways to), best practices.*?, (search|find|discover).*?(information|examples|docs)

From: skill-rules.json - research-trigger configuration

Purpose

Automatically trigger the /research slash command when users ask questions requiring deep research, best practices lookup, or multi-source information gathering using natural language. Eliminates the need to remember the /research syntax.

Core Behavior

In any DAIC mode:

  1. Question Detection

    • Detect when user asks research-oriented questions
    • Examples: "How do I implement JWT auth?", "What are best practices for caching?"
    • Automatically invoke /research with the question

  2. Intent Classification

    • Identify research intent patterns:
      • How-to questions ("How do I...")
      • Comparison questions ("What are the options for...")
      • Best practices queries ("Best practices for...")
      • Investigation requests ("Research X", "Investigate Y")

  3. Research Invocation

    • Call /research command with extracted question
    • Pass through to parallel research agents
    • Return comprehensive research findings with citations

Natural Language Examples

Triggers this skill:

  • ✓ "Research best practices for error handling"
  • ✓ "How do I implement OAuth2?"
  • ✓ "What are the options for state management?"
  • ✓ "Find information about GraphQL subscriptions"
  • ✓ "Investigate performance optimization techniques"
  • ✓ "Look up JWT token refresh patterns"
  • ✓ "Search for examples of Sentry integration"
  • ✓ "Discover alternatives to Express middleware"
  • ✓ "What are best practices for API versioning?"

Does NOT trigger:

  • ✗ "What is the current mode?" (DAIC mode question, not research)
  • ✗ "What's in progress?" (project status, not research)
  • ✗ "What files changed?" (git status, not research)

Safety Guardrails

ANALYSIS-ONLY RULES:

  • ✓ NEVER call write tools (Edit, Write, MultiEdit)
  • ✓ Only invokes /research command which is read-only
  • ✓ Safe to run in any DAIC mode
  • ✓ Provides research findings and recommendations only

Invocation Rules:

  • Only invoke /research when clear research intent detected
  • Extract the full question from user's message
  • Handle "how", "what", "why" questions appropriately
  • Distinguish research questions from status/info queries

Implementation Notes

This skill acts as a natural language wrapper around /research. The actual research logic lives in .claude/commands/research.md, which:

  • Classifies query type (breadth-first vs depth-first)
  • Launches parallel research subagents
  • Synthesizes findings with automatic citations
  • Returns comprehensive research report

This skill simply detects the intent and routes to that command.