Bilibili 热门趋势分析

获取 Bilibili 全榜单热门数据并分析趋势。支持 21 个榜单，自动调用子 Agent 分析并生成 MD 报告持久化储存。安全无隐私风险，仅调用公开 API。

Security Scan

The skill's code, instructions, and requested resources are coherent with its stated purpose (fetch public Bilibili ranking data, analyze, and store reports locally); it does not request credentials or perform unexpected external network calls beyond Bilibili's public APIs and the platform's subagent API.

Purpose & Capability

Name/description match code behavior: scripts call Bilibili public ranking APIs, compute stats/keywords, generate prompts and save JSON/MD reports. No unrelated binaries, credentials, or external services are requested.

Instruction Scope

SKILL.md instructs running the included Python scripts and describes local storage and using OpenClaw subagents. The runtime instructions and code reference only workspace paths and Bilibili public APIs. There are no instructions to read unrelated system files or to exfiltrate secrets.

Install Mechanism

No install spec; this is an instruction+script package. All code is included in the bundle; there are no downloads or archive extracts from remote URLs. Risk from installation is minimal.

Credentials

The skill requires no environment variables, credentials, or config paths. Network access is limited to Bilibili public API endpoints used for the stated function. The use of the platform's sessions_spawn (if available) is the only non-Bilibili interaction and is justified by the described subagent analysis feature.

Persistence & Privilege

always:false (normal). The scripts may autonomously call sessions_spawn to run a subagent and send the generated prompt + dataset for analysis — this is expected for automatic analysis but means fetched data is passed to the platform's subagent layer. If you want to avoid that, the skill provides a manual mode (--manual) and the SKILL.md documents local storage.

Assessment

This skill appears to do what it claims: fetch public Bilibili ranking data, analyze it, and save local reports. Before installing or running it: 1) If you are sensitive about sending collected data to platform subagents, run with --manual or inspect/disable sessions_spawn on your environment — the code will attempt to call sessions_spawn to run a child agent and will send the analysis prompt and full data. 2) The scripts create {workspace}/json and {workspace}/memory/bilibili-analysis and write JSON/MD files there — ensure that is acceptable in your environment. 3) The code sleeps briefly between requests but may still trigger Bilibili rate limits; use lower frequency if you run it repeatedly. 4) No credentials or external downloads are requested, and no attempts to read other system credentials were found. If you want additional assurance, review the sessions_spawn implementation on your platform (to confirm where prompts/data are routed) or run the scripts in an isolated workspace/container.

Bilibili Trending

获取 B 站热门数据 → 自动调用子 Agent 分析 → 持久化储存 → 趋势预测

安全说明

• 仅调用公开 API：使用 B 站公开排行榜 API，不涉及登录、Cookie、用户信息
• 无个人信息：只获取热门视频数据，不采集任何用户身份信息
• 本地存储：数据保存在本地工作区，不上传到任何外部服务器
• 标准请求头：仅使用标准 User-Agent，无特殊权限
• 注意：频繁请求可能触发 API 限流（-352 错误），建议降低请求频率

环境要求

• Python 3.x
• requests 库
• OpenClaw 环境（用于自动调用子 Agent）

支持 21 个榜单

类型	榜单
普通视频	全站、动画、游戏、音乐、舞蹈、鬼畜、影视、娱乐、知识、科技数码、美食、汽车、时尚美妆、体育运动、动物
PGC 内容	番剧、国创、纪录片、电影、电视剧

---

使用方式

以下命令在 skill 目录下执行（脚本会根据相对于工作区的位置自动创建所需目录）

抓取并分析单个榜单

cd skills/Bilibili-trending/scripts
python bilibili_all.py --rank <榜单>

示例：

python bilibili_all.py --rank game    # 游戏榜
python bilibili_all.py --rank tv      # 电视剧榜
python bilibili_all.py --rank anime   # 番剧榜
python bilibili_all.py --rank all     # 全站榜

列出所有榜单

python bilibili_all.py --list

手动模式（不调用子 Agent）

python bilibili_all.py --rank game --manual

---

完整流程

Step 1: 抓取数据

脚本自动完成：

1. 调用 B 站 API 抓取数据
2. 处理数据（计算互动率、提取关键词）
3. 保存 JSON 到 {工作区}/json/output_{rank_type}.json
4. 更新趋势数据到 trend.json

Step 2: 自动分析

脚本自动 spawn 子 Agent，发送分析 prompt

Step 3: 保存报告

分析完成后，报告自动保存到：

{工作区}/memory/bilibili-analysis/{榜单名称}_{时间}.md

例如：游戏_2026-04-02-15-30-45.md

---

趋势分析命令

在 skills/Bilibili-trending/scripts 目录下执行

查看趋势

python bili_trend.py trend          # 全局趋势
python bili_trend.py trend game     # 单榜单趋势

生成周总结

python bili_trend.py weekly         # 全站周总结
python bili_trend.py weekly game    # 单榜单周总结

生成月总结

python bili_trend.py monthly        # 全站月总结
python bili_trend.py monthly game   # 单榜单月总结

---

非 OpenClaw 环境

如果没有 OpenClaw 环境：

• 脚本检测到无法导入 sessions_spawn
• 自动输出 prompt 供手动使用
• 用户可将 prompt 发送给子 Agent 手动分析

---

关键词提取逻辑

# 1. 正则提取 2-4 字中文
words = re.findall(r'[\u4e00-\u9fa5]{2,4}', title)

# 2. 统计词频
kw_counter = Counter(words)

# 3. 取 Top 5
top_keywords = [kw for kw, _ in kw_counter.most_common(5)]

---

输出文件

脚本会在工作区自动创建以下目录结构：

{工作区}/
├── json/                           # JSON 数据目录
│   └── output_{rank_type}.json      # 原始数据
└── memory/bilibili-analysis/        # 分析结果目录
    ├── trend.json                    # 趋势累计数据
    ├── 游戏_2026-04-02-15-30-45.md  # 分析报告
    ├── weekly-2026-W14.md           # 周总结
    └── monthly-2026-04.md          # 月总结

---

注意事项

• 频繁请求会触发 API 限流（-352 错误），建议降低请求频率或等待后重试
• 趋势数据需长期积累（建议 30+ 次）才能形成可靠预测
• 子 Agent 分析完成后报告自动保存