Search Hierarchy

This enforces a token efficient search strategy by routing queries to the right tool. It pushes you toward AST-grep for structural patterns like function definitions (around 50 tokens), LEANN for semantic questions about how systems work (around 100 tokens), and reserves grep for exact identifier searches. The hook system actually blocks inefficient tool choices and suggests alternatives, which is aggressive but probably necessary since the default behavior wastes thousands of tokens reading files you don't need. The decision tree is sensible: find files first with cheap searches, then read only what matters. Works best if you're dealing with token budgets on large codebases where a few wrong reads blow your context window.

skills.sh installs: 478
GitHub stars: 3.9K
last updated: Aug 21, 2026

View on GitHub

Install

$ npx -y skills add parcadei/continuous-claude-v3 --skill search-hierarchy --agent claude-code

Installs into .claude/skills of the current project.

Files

Search Tool Hierarchy

Use the most token-efficient search tool for each query type.

Decision Tree

Query Type?
├── STRUCTURAL (code patterns)
│   → AST-grep (~50 tokens output)
│   Examples: "def foo", "class Bar", "import X", "@decorator"
│
├── SEMANTIC (conceptual questions)
│   → LEANN (~100 tokens if path-only)
│   Examples: "how does auth work", "find error handling patterns"
│
├── LITERAL (exact identifiers)
│   → Grep (variable output)
│   Examples: "TemporalMemory", "check_evocation", regex patterns
│
└── FULL CONTEXT (need complete understanding)
    → Read (1500+ tokens)
    Last resort after finding the right file

Token Efficiency Comparison

Tool	Output Size	Best For
AST-grep	~50 tokens	Function/class definitions, imports, decorators
LEANN	~100 tokens	Conceptual questions, architecture, patterns
Grep	~200-2000	Exact identifiers, regex, file paths
Read	~1500+	Full understanding after finding the file