code-search-selector

The code-search-selector skill serves as an intelligent decision-making layer for development workflows, designed to guide users toward the most effective search method for their specific task. By evaluating the intent behind a user's request, the skill determines whether a semantic search using claudemem is superior for high-level conceptual questions or if a native tool like Grep or Glob is more appropriate for exact pattern matching. This workflow optimization reduces unnecessary token consumption while ensuring high-quality, ranked search results.

• Automatically differentiates between conceptual queries such as 'how does X work' or 'audit integration' and exact string or symbol lookups.

• Provides a proactive decision tree that checks claudemem indexing status before execution, offering to index repositories when necessary to guarantee semantic accuracy.

• Facilitates advanced semantic capabilities like architecture mapping via 'claudemem --agent map', data flow tracing, and implementation discovery across large codebases.

• Integrates with native command-line utilities for scenarios requiring exact matches, such as finding specific flag constants, counting occurrences of TODOs, or filtering by file pattern.

• Features a fallback mechanism that routes complex investigative tasks to the code-analysis:detective agent when semantic indices are unavailable.

• Delivers ranked, relevant code chunks using PageRank and relevance scoring, which significantly outperforms standard raw text matching for large projects.

• Always perform a 'claudemem status' check before attempting semantic queries; if the index is missing, initiate 'claudemem index -y' to ensure maximum performance.

• Use semantic search for: architectural investigations, tracing data flow, finding implementations by concept, auditing API usage, and broad feature exploration.

• Use native Grep for: exact string matches, regex patterns, counting symbol occurrences, and simple file pattern filters where relevance ranking is not required.

• When the hook system provides proactive claudemem results, prioritize these over manual native tool searches to benefit from relevance ranking and token efficiency.

• The tool is designed for software engineers and architects managing complex, evolving codebases who need to balance speed, cost (token usage), and depth of analysis.