Explain - Deep Code Explanation

Get a comprehensive explanation of code, files, directories, or architectural concepts. Automatically dispatches a general-purpose agent preloaded with the most relevant -ops skill and uses modern CLI tools for analysis.

Arguments

$ARGUMENTS

• <target> - File path, function name, class name, directory, or concept
• --depth <shallow|normal|deep|trace> - Level of detail (default: normal)
• --focus <arch|flow|deps|api|perf> - Specific focus area

Architecture

/explain <target> [--depth] [--focus]
    |
    +-> Step 1: Detect & Classify Target
    |     +- File exists? -> Read it
    |     +- Function/class? -> ast-grep to find definition
    |     +- Directory? -> tokei for overview
    |     +- Concept? -> rg search codebase
    |
    +-> Step 2: Gather Context (parallel)
    |     +- structural-search skill -> find usages
    |     +- code-stats skill -> assess scope
    |     +- Find related: tests, types, docs
    |     +- Load: AGENTS.md, CLAUDE.md conventions
    |
    +-> Step 3: Route to Explainer (general-purpose + skill preload)
    |     +- .ts/.tsx -> general-purpose, preload typescript-ops or react-ops
    |     +- .py -> general-purpose, preload python-* skill by topic
    |     +- .vue -> general-purpose, preload vue-ops
    |     +- .sql/migrations -> general-purpose, preload postgres-ops
    |     +- agents/skills/commands -> general-purpose, preload claude-code-ops
    |     +- Default -> general-purpose
    |     +- All explainers preload: debug-ops (systematic analysis)
    |
    +-> Step 4: Generate Explanation
    |     +- Structured markdown with sections
    |     +- Mermaid diagrams (flowchart/sequence/class)
    |     +- Related code paths as file:line refs
    |     +- Design decisions and rationale
    |
    +-> Step 5: Integrate
          +- Offer to save to ARCHITECTURE.md (if significant)
          +- Link to /save if working on related task

Execution Steps

Step 1: Detect Target Type

# Check if target is a file
test -f "$TARGET" && echo "FILE" && exit

# Check if target is a directory
test -d "$TARGET" && echo "DIRECTORY" && exit

# Otherwise, search for it as a symbol

For files: Read directly with bat (syntax highlighted) or Read tool.

For directories: Get overview with tokei (if available):

command -v tokei >/dev/null 2>&1 && tokei "$TARGET" --compact || echo "tokei unavailable"

For symbols (function/class): Find definition with ast-grep:

# Try ast-grep first (structural)
command -v ast-grep >/dev/null 2>&1 && ast-grep -p "function $TARGET" -p "class $TARGET" -p "def $TARGET"

# Fallback to ripgrep
rg "(?:function|class|def|const|let|var)\s+$TARGET" --type-add 'code:*.{ts,tsx,js,jsx,py,vue}' -t code

Step 2: Gather Context

Run these in parallel where possible:

Find usages (structural-search skill):

# With ast-grep
ast-grep -p "$TARGET($_)" --json 2>/dev/null | head -20

# Fallback
rg "$TARGET" --type-add 'code:*.{ts,tsx,js,jsx,py,vue}' -t code -l

Find related files:

# Tests
fd -e test.ts -e spec.ts -e test.py -e spec.py | xargs rg -l "$TARGET" 2>/dev/null

# Types/interfaces
fd -e d.ts -e types.ts | xargs rg -l "$TARGET" 2>/dev/null

Load project conventions:

• Read AGENTS.md if exists
• Read CLAUDE.md if exists
• Check for framework-specific patterns

Step 3: Route to Explainer

Dispatch is skills-first: the generic general-purpose subagent preloads the relevant -ops skill based on file extension and content:

Pattern	Dispatch	Preload	Condition
.ts	general-purpose	skills/typescript-ops/SKILL.md	No JSX/React imports
.tsx	general-purpose	skills/react-ops/SKILL.md	JSX present
.js, .jsx	general-purpose	skills/javascript-ops/SKILL.md	-
.py	general-purpose	relevant skills/python-*/SKILL.md by topic	-
.vue	general-purpose	skills/vue-ops/SKILL.md	-
.sql, migrations/*	general-purpose	skills/postgres-ops/SKILL.md	-
agents/*.md, skills/*, commands/*	general-purpose	skills/claude-code-ops/SKILL.md	Claude extensions
*.test.*, *.spec.*	general-purpose	(framework skill by file type)	-
Other	general-purpose	-	Fallback

Invoke via Task tool:

Task tool with subagent_type: "general-purpose" (preload claude-code-ops for Claude extensions)
model: "sonnet"
Prompt includes:
  - Skill preloading (domain knowledge):
    "First, read these files for analysis context:
     - Read: skills/debug-ops/SKILL.md
     - Read: [Preload column for the matched pattern]"
  - File content
  - Related files found
  - Project conventions
  - Requested depth and focus

Step 4: Generate Explanation

The dispatched agent produces a structured explanation:

# Explanation: [target]

## Overview
[1-2 sentence summary of purpose and role in the system]

## Architecture

[Mermaid diagram - choose appropriate type]

### Flowchart (for control flow)
` ` `mermaid
flowchart TD
    A[Input] --> B{Validate}
    B -->|Valid| C[Process]
    B -->|Invalid| D[Error]
    C --> E[Output]
` ` `

### Sequence (for interactions)
` ` `mermaid
sequenceDiagram
    participant Client
    participant Server
    participant Database
    Client->>Server: Request
    Server->>Database: Query
    Database-->>Server: Result
    Server-->>Client: Response
` ` `

### Class (for structures)
` ` `mermaid
classDiagram
    class Component {
        +props: Props
        +state: State
        +render(): JSX
    }
` ` `

## How It Works

### Step 1: [Phase Name]
[Explanation with code references]

See: `src/module.ts:42`

### Step 2: [Phase Name]
[Explanation]

## Key Concepts

### [Concept 1]
[Explanation]

### [Concept 2]
[Explanation]

## Dependencies

| Import | Purpose |
|--------|---------|
| `package` | [why it's used] |

## Design Decisions

### Why [decision]?
[Rationale and tradeoffs considered]

## Related Code

| File | Relationship |
|------|--------------|
| `path/to/file.ts:123` | [how it relates] |

## See Also

- `/explain path/to/related` - [description]
- [External docs link] - [description]

Depth Modes

Mode	Output
--shallow	Overview paragraph, key exports, no diagram
--normal	Full explanation with 1 diagram, main concepts (default)
--deep	Exhaustive: all internals, edge cases, history, multiple diagrams
--trace	Data flow tracing through entire system, sequence diagrams

Shallow Example

/explain src/auth/token.ts --shallow

Output: Single paragraph + exports list.

Deep Example

/explain src/core/engine.ts --deep

Output: Full internals, algorithm analysis, performance notes, edge cases.

Trace Example

/explain handleLogin --trace

Output: Traces data flow from entry to database to response.

Focus Modes

Mode	What It Analyzes
--focus arch	Module boundaries, layer separation, dependencies
--focus flow	Data flow, control flow, state changes
--focus deps	Imports, external dependencies, integrations
--focus api	Public interface, inputs/outputs, contracts
--focus perf	Complexity, bottlenecks, optimization opportunities

CLI Tool Integration

Commands use modern CLI tools with graceful fallbacks:

Tool	Purpose	Fallback
tokei	Code statistics	Skip stats
ast-grep	Structural search	rg with patterns
bat	Syntax highlighting	Read tool
rg	Content search	Grep tool
fd	File finding	Glob tool

Check availability:

command -v tokei >/dev/null 2>&1 || echo "tokei not installed - skipping stats"

Usage Examples

# Explain a file
/explain src/auth/oauth.ts

# Explain a function (finds it automatically)
/explain validateToken

# Explain a directory
/explain src/services/

# Deep dive with architecture focus
/explain src/core/engine.ts --deep --focus arch

# Trace data flow
/explain handleUserLogin --trace

# Quick overview
/explain src/utils/helpers.ts --shallow

# Focus on dependencies
/explain package.json --focus deps

Integration

Skill/Command	Relationship
/review	Review after understanding
/testgen	Generate tests for explained code
/save	Save progress if working on related task

Persistence

After significant explanations, you may be offered:

Would you like to save this explanation?
  1. Append to ARCHITECTURE.md
  2. Append to AGENTS.md (if conventions-related)
  3. Don't save (output only)

This keeps valuable architectural knowledge in git-tracked documentation.

Notes

• Explanations are based on code analysis, not documentation
• Complex systems may need multiple /explain calls
• Use --deep for unfamiliar codebases
• Mermaid diagrams render in GitHub, GitLab, VSCode, and most markdown viewers
• Preloaded -ops skills provide framework-specific insights