Provide comprehensive guidance on Claude Code's native tools. These tools are optimized for agent use and should ALWAYS be preferred over equivalent bash commands.
NEVER Use Bash For These Operations
| Bad (Bash) |
Good (Native) |
Why |
find . -name "*.ts" |
Glob({pattern: "**/*.ts"}) |
Faster, respects gitignore |
grep -r "pattern" . |
Grep({pattern: "pattern"}) |
Structured output, pagination |
cat file.txt |
Read({file_path: "file.txt"}) |
Line numbers, handles images/PDFs |
head -n 50 file.txt |
Read({file_path: "file.txt", limit: 50}) |
Native pagination |
sed -i 's/old/new/' |
Edit({old_string: "old", new_string: "new"}) |
Safe, atomic, validated |
echo "text" > file |
Write({file_path: "file", content: "text"}) |
Permission checked |
Tool Categories
| Category |
Reference |
Tools |
| File Search |
references/file-search.md |
Glob, Grep |
| Structural Search |
See ast-grep skill |
ast-grep (sg) for AST patterns |
| File Operations |
references/file-operations.md |
Read, Edit, Write |
| User Interaction |
references/user-interaction.md |
AskUserQuestion, TodoWrite |
| Agent Orchestration |
references/orchestration.md |
Task, TaskOutput |
| External Resources |
references/external.md |
WebFetch, WebSearch, Bash |
| MCP Tools |
references/external.md |
Context7, OctoCode, Codex |
Skill References
| Skill |
When to Use |
ast-grep |
Structural code patterns (function calls, imports, types) |
Tool Permission Matrix
| Tool |
Permission Required |
Description |
| Glob |
No |
Find files by pattern |
| Grep |
No |
Search file contents |
| Read |
No |
Read file contents |
| Edit |
Yes |
Modify files |
| Write |
Yes |
Create/overwrite files |
| AskUserQuestion |
No |
Interactive questions |
| TodoWrite |
No |
Task tracking |
| Task |
No |
Spawn sub-agents |
| TaskOutput |
No |
Get agent results |
| Bash |
Yes |
Shell commands |
| WebFetch |
Yes |
Fetch URLs |
| WebSearch |
Yes |
Web search |
| NotebookEdit |
Yes |
Edit Jupyter notebooks |
Search Then Read
// Find files matching pattern
Glob({pattern: "src/**/*.ts"})
// Search for content (text patterns)
Grep({pattern: "function handleAuth", type: "ts"})
// Read specific file
Read({file_path: "src/auth/handler.ts"})
Structural Code Search (ast-grep skill)
For complex code patterns, use ast-grep instead of Grep:
# Find all async functions
sg -p "async function $FUNC($$$)" -l typescript
# Find React hooks
sg -p "use$HOOK($$$)" -l tsx
# Find specific imports
sg -p 'import { $$$IMPORTS } from "react"' -l typescript
Documentation Lookup (MCP skill)
Prefer MCP tools over WebFetch for library docs:
// Context7 for library documentation
mcp__plugin_crew_context7__resolve-library-id({libraryName: "react-query"})
mcp__plugin_crew_context7__query-docs({
context7CompatibleLibraryID: "/tanstack/query",
topic: "query invalidation"
})
// OctoCode for GitHub code search
mcp__plugin_crew_octocode__githubSearchCode({
keywordsToSearch: ["useQuery"],
owner: "tanstack",
repo: "query"
})
// Codex for deep reasoning
mcp__plugin_crew_codex__codex({
prompt: "Analyze architectural trade-offs...",
sandbox: "read-only"
})
Edit Workflow
// ALWAYS read before edit
Read({file_path: "path/to/file.ts"})
// Make targeted edit
Edit({
file_path: "path/to/file.ts",
old_string: "const x = 1",
new_string: "const x = 2"
})
User Decision Points
AskUserQuestion({
questions: [{
question: "Which approach should we use?",
header: "Approach",
options: [
{label: "Option A (Recommended)", description: "Fastest, simplest"},
{label: "Option B", description: "More flexible"},
{label: "Option C", description: "Most robust"}
],
multiSelect: false
}]
})
Agent Orchestration
// Launch parallel agents
Task({
subagent_type: "Explore",
prompt: "Find all API endpoints in src/",
description: "Find API endpoints",
run_in_background: true
})
// Retrieve results
TaskOutput({task_id: "abc123", block: true})
majiayu000