360 Secure SkillHub
360 安全技能中心

AI Code Review

核心功能涵盖GPT-4o代码差异分析、语音评审记录转录、Discord通知推送,支持全流程拉取请求审查;具备URL安全校验、请求超时控制等保障机制,提供标准化API接口。适用于开发团队自动化代码评审,输入含PR编号、差异URL等,输出结构化审查结果及可选转录文本,提升评审效率。

terrycarter1985
terrycarter1985
热度 40.76
最近更新 2026-05-12 07:56:51
分类:开发依赖:代码执行 / 读写文件 / 第三方 API应用场景:代码审查 / 自动化脚本 / API 集成

安装技能

复制下面的话术发给你的 Agent,它会通过 secure-skills 检索并安装该技能。

执行命令安装此 skill: npx secure-skills-cli add ai-code-review-svc
Metadata
name
ai-code-review
version
1.1.0
description
Automated code review with LLM analysis, voice transcription, and Discord notifications
author
openclaw
tags
["code-review","llm","whisper","discord","automation"]

AI Code Review Skill

Automated code review service that combines LLM-powered diff analysis, voice-note transcription, and Discord notifications into a unified review pipeline.

Features

  • LLM Code Analysis: Sends diffs to GPT-4o with structured JSON output (issues count, suggestions, approval status, summary)
  • Voice Note Transcription: Transcribes review meeting recordings via OpenAI Whisper API with file validation
  • Discord Notifications: Rich embed notifications with approval status, issue counts, and color-coded indicators
  • URL Safety: Scheme validation prevents SSRF attacks on all fetched URLs
  • Request Timeouts: All HTTP calls enforce a 30-second timeout to prevent hanging
  • Structured Logging: Consistent logging throughout with logging module
  • ClawHub Publishing: Built-in publish_skill() helper using the ClawHub CLI

Quick Start

# Set required environment variables
export OPENAI_API_KEY="sk-..."
export DISCORD_WEBHOOK_URL="https://discord.com/api/webhooks/..."  # optional

# Run a review
python src/code_review_service.py

API Reference

process_pull_request(pr_number, diff_url, voice_note_path=None)

End-to-end PR review: fetches diff, runs LLM analysis, optionally transcribes a voice note, and sends a Discord notification.

Returns a dict with keys: issues_found, suggestions, approval, summary, and optionally voice_note_transcription.

analyze_code_changes(diff_content)

Sends diff text to GPT-4o for analysis. Returns structured JSON with issues_found, suggestions, approval (approved/needs_changes/rejected), and summary.

transcribe_voice_note(audio_file_path)

Validates the audio file exists and is non-empty, then transcribes via Whisper-1. Returns the transcription text.

send_discord_notification(message, embed=None)

Posts a message (with optional rich embed) to the configured Discord webhook. Returns True on success.

publish_skill(skill_path, version)

Publishes a skill directory to ClawHub at the given version using the clawhub CLI.

Configuration

Variable Required Default Description
OPENAI_API_KEY Yes — OpenAI API key for GPT-4o and Whisper
DISCORD_WEBHOOK_URL No — Discord webhook URL for notifications
CLAWHUB_API_URL No https://api.clawhub.com/v1 ClawHub API base URL

Health Check

The included scripts/healthcheck.sh monitors nginx, docker, code-review-service, and whisper-api-gateway. It auto-restarts failed services and sends Discord alerts. Disk usage warnings trigger at 80% and critical alerts at 90%.

Changelog

1.1.0

  • Fixed SSRF vulnerability: added _validate_url() with scheme allowlist for all fetched URLs
  • Added 30-second request timeouts to all requests calls (diff fetch + Discord webhook)
  • Replaced bare except clauses with specific exception types (requests.RequestException, FileNotFoundError, ValueError, json.JSONDecodeError)
  • Added input validation: file existence/size checks for audio, empty-diff handling
  • Implemented actual LLM-based code analysis via GPT-4o (replaced stub analyze_code_changes)
  • Added logging module throughout; removed silent error swallowing
  • Lazy-initialized OpenAI client with clear error on missing key
  • Improved error handling in voice note transcription (graceful skip on failure)
  • Enhanced health check script compatibility with chroot/container environments