What problem does the Code Search Selector solve?

It prevents the AI from defaulting to inefficient 'grep' or bulk 'read' commands for tasks that require semantic understanding, ensuring more accurate and context-aware results.

Can I still use standard Grep for exact matches?

Yes. The skill includes a decision tree that specifically allows the use of Grep for finding exact strings, counting occurrences, or matching regex patterns.

Do I need claudemem installed to use this skill?

Yes, this skill is optimized to leverage claudemem for semantic indexing, though it provides a decision tree to handle cases where an index hasn't been created yet.

When does this skill automatically trigger?

The skill auto-invokes when keywords like 'audit', 'trace', or 'how does it work' are used, or when the system detects an attempt to read 3 or more files sequentially to understand a feature.

How does this improve performance during coding sessions?

By using semantic search instead of sequential reads, it retrieves relevant code chunks faster and significantly reduces token consumption.

Semantic Code Search Selector

Name: Semantic Code Search Selector
Author: involvex

byinvolvex

开发者工具

Optimizes codebase investigations by enforcing semantic search over manual file reads and grep operations.

关于

The Semantic Code Search Selector is a mandatory gatekeeping skill designed to eliminate tool familiarity bias during complex code analysis. It intercepts broad queries—such as those involving architectural mapping, data flow tracing, or comprehensive audits—and redirects the AI from using shallow string-matching tools like grep toward deep semantic search via claudemem. By ensuring the AI uses the right tool for the job, this skill provides higher-quality context, reduces token overhead from bulk file reading, and significantly improves the accuracy of Claude’s codebase investigations.

主要功能

Intelligent routing to specialized architect, developer, or debugger subagents
Mandatory decision tree for selecting the optimal search tool based on intent
Automatic interception of semantic investigation queries like 'audit' or 'trace'
0 GitHub stars
Prevents inefficient bulk file reads and broad glob pattern overhead
Direct integration with claudemem for high-speed semantic indexing

使用场景

Mapping high-level architecture and patterns in unfamiliar codebases
Tracing complex data flows and logic across multiple service integration points
Conducting thorough security or logic audits for specific application features

关于

主要功能

Intelligent routing to specialized architect, developer, or debugger subagents
Mandatory decision tree for selecting the optimal search tool based on intent
Automatic interception of semantic investigation queries like 'audit' or 'trace'
0 GitHub stars
Prevents inefficient bulk file reads and broad glob pattern overhead
Direct integration with claudemem for high-speed semantic indexing

使用场景

Mapping high-level architecture and patterns in unfamiliar codebases
Tracing complex data flows and logic across multiple service integration points
Conducting thorough security or logic audits for specific application features