codeflow-analyzer

Code Flow Analyzer

Systematically trace the complete call chain of any feature or process in a codebase, from entry point to the deepest dependency, and produce a structured analysis document.

When This Skill Applies

• "这个功能的调用链是怎样的？"
• "Trace the request flow from X to Y"
• "How does feature Z work under the hood?"
• "Map the data flow of this process"
• "I want to understand the full path from user input to final output"
• Any question about how layers connect: frontend → API → service → domain → storage

Core Approach: Layer-by-Layer Tracing

Real-world codebases are layered. Each layer has its own conventions for entry points, routing, and delegation. Instead of guessing the architecture, discover it empirically by following the actual code paths.

Phase 1: Identify Entry Point

Start from the user-facing trigger:

1. Find the surface-level entry — UI component, CLI command, API endpoint, message handler, cron trigger, etc.
2. Identify what it calls — the first cross-boundary call (HTTP request, function import, message publish, RPC call).

Techniques:

• Grep for keywords from the user's description (feature name, button label, route path, event name)
• Glob for files matching the feature name pattern
• Read the entry file to find outbound calls

Phase 2: Trace Through Each Layer

For each layer discovered, answer three questions:

1. How does this layer receive the call? (route handler, message consumer, function signature)
2. What does this layer do with it? (validation, transformation, delegation, side effects)
3. Where does it hand off next? (outbound call, return value, event emission)

Repeat until you reach a terminal point (database write, external API call, file output, response return).

Parallel exploration strategy: When entering a new layer, launch multiple searches in parallel:

• Grep for the function/class/route name being called
• Glob for files matching the module/package name
• Read configuration files that might define routing or wiring