AgentSkillsCN

codex-explain

借助 Codex MCP 解释复杂的代码逻辑。支持简要、常规与深度三种不同层次的解读。

SKILL.md
--- frontmatter
name: codex-explain
description: Explain complex code logic using Codex MCP. Supports brief/normal/deep depth levels.
allowed-tools: mcp__codex__codex, mcp__codex__codex-reply, Read, Grep, Glob

Codex Explain Skill

Trigger

  • •Keywords: explain code, what does this do, how does this work, code walkthrough

When NOT to Use

  • •Code review (use codex-code-review)
  • •Bug investigation (use bug-fix or issue-analyze)
  • •Architecture overview (use code-explore)

Workflow

code
Read target → [Collect context] → Codex explain → Output explanation

Step 1: Read Target File

Read file content. If --lines specified, extract only that range.

Step 2: Codex Explanation

Use mcp__codex__codex with explanation prompt. See @references/codex-prompt-explain.md.

Config: sandbox: 'read-only', approval-policy: 'never'

Depth Levels

LevelDescription
briefOne-sentence summary
normalFunctional overview + execution flow + key concepts (default)
deep+ Design patterns + complexity + issues + dependencies

Verification

  • • Codex independently researched project context (imports, callers)
  • • Explanation matches requested depth level
  • • Key concepts are identified and explained

References

  • •Explanation prompt: references/codex-prompt-explain.md

Examples

code
Input: /codex-explain src/service/order/order.service.ts
Action: Read file → Codex explain (normal) → Functional summary + detailed explanation

Input: /codex-explain src/service/xxx.ts --depth deep
Action: Read file → Codex explain (deep) → Patterns + complexity + issues

Input: /codex-explain src/xxx.ts --lines 50-100
Action: Read lines 50-100 → Codex explain → Focused explanation