Does this skill stop me from reading files?

No, it simply ensures you don't use 'Read' as a brute-force method for understanding a feature when a single semantic query would be more efficient.

When does the Code Search Selector skill trigger?

It auto-triggers on keywords like 'audit', 'investigate', or 'how does X work', and whenever the agent attempts to read 3 or more files or use broad Glob patterns.

Do I need claudemem for this skill to work?

While the skill works best with an indexed claudemem environment, it includes fallbacks to specialized detective sub-agents if an index is not available.

Why should I use semantic search instead of grep?

Grep is ideal for exact string matches, but semantic search understands the conceptual meaning of code, making it far superior for architectural questions or tracing complex logic.

Code Search Selector

Name: Code Search Selector
Author: tianzecn

bytianzecn

•

开发者工具

Optimizes codebase exploration by intercepting broad file reads and redirecting to high-accuracy semantic search.

关于

The Code Search Selector skill acts as an intelligent gatekeeper for codebase investigations, preventing inefficient 'native tool' habits like bulk file reading or broad grepping. It automatically triggers during audits, architectural mappings, and feature investigations to ensure Claude uses semantic search via claudemem for superior context and faster results. By classifying tasks into semantic or exact-match categories, it ensures the most efficient tool is used for the job, drastically reducing token waste and improving the accuracy of complex code analysis.

主要功能

Automatic keyword interception for 'audit', 'investigate', and 'trace' queries
Interception of broad Glob patterns and bulk sequential file reads to save tokens
Seamless integration with claudemem for indexed semantic exploration
Direct hand-off to specialized detective sub-agents for deep analysis
1 GitHub stars
Mandatory decision tree to distinguish between semantic search and exact matching

使用场景

Tracing feature implementations without manually reading dozens of files
Mapping high-level architecture and data flows in unfamiliar codebases
Conducting comprehensive logic audits across multiple integration points

关于

主要功能

Automatic keyword interception for 'audit', 'investigate', and 'trace' queries
Interception of broad Glob patterns and bulk sequential file reads to save tokens
Seamless integration with claudemem for indexed semantic exploration
Direct hand-off to specialized detective sub-agents for deep analysis
1 GitHub stars
Mandatory decision tree to distinguish between semantic search and exact matching

使用场景

Tracing feature implementations without manually reading dozens of files
Mapping high-level architecture and data flows in unfamiliar codebases
Conducting comprehensive logic audits across multiple integration points