AgentSkillsCN

deep-research

一套基于向量的科研技能，专为解决复杂的交通场景而设计（如无障碍出行、情绪匹配、最后一公里接驳等）。

SKILL.md
--- frontmatter
name: deep-research
description: A suite of vector-powered research skills for solving complex transit scenarios (Accessibility, Vibe Matching, Last Mile).
allowed-tools:
  - mcp_supabase-mcp-server_execute_sql
  - mcp_supabase-mcp-server_search_docs
tags:
  - deep-research
  - strategy
  - vector-search

Deep Research Skills

This skill set enables the agent to perform "Deep Research" into specific transit domains using semantic search and vector matching.

Included Strategies

StrategyGoalFile
Vibe MatcherFind places with similar atmosphere but less crowded.reference/vibe-matcher.md
Facility PathfinderDetailed vertical navigation for stroller/wheelchair.reference/facility-pathfinder.md
Last Mile ConnectorSolve the "Station to Final Destination" gap (>1km).reference/last-mile-connector.md
Spatial ReasonerCalculate alternative routes during train suspension.reference/spatial-reasoner.md

Usage Principles

  • •Vector First: These skills rely on vibe_embedding or facility graph data, necessitating vector search or specialized graph queries.
  • •Prompt Engineering: Each strategy defines specific JSON output formats and persona tones (e.g., "Guardian" for accessibility).