Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add asaotomo/DeepSentry --skill bilibili-playgit clone --depth 1 https://github.com/asaotomo/DeepSentryWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/asaotomo/deepsentry/bilibili-play)<a href="https://agentmods.dev/skills/asaotomo/deepsentry/bilibili-play"><img src="https://agentmods.dev/badge/skills/asaotomo/deepsentry/bilibili-play/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/asaotomo/deepsentry/bilibili-play"><img src="https://agentmods.dev/badge/skills/asaotomo/deepsentry/bilibili-play.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- NVIDIA SkillSpector pass
What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00100 | $0.01779 |
| Opus 5.5 | $0.00040 | $0.00712 |
| Sonnet 5.5 | $0.00020 | $0.00356 |
| Haiku 4.5 | $0.00010 | $0.00178 |
Grade A, and why
bilibili-play scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 2mo ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 142 lines — stays where its author put it; the contents beside it link to each section on GitHub.
在真实浏览器播放 B站视频(倍速 / 全屏 / 抓帧)
这是 HawkEye 真实浏览器直播放 工作流,不是下载,也不是内置 browser_browse。
HawkEye 已出现在本会话工具列表中就等于已经连通,禁止再做 MCP 安装/端口/MD5 自检。
工具名以本会话实际前缀为准(常见 hx0-hawkeye__browser_navigate)。下文写短名。
何时使用
用户要在浏览器里 看/播放 B站视频,尤其带全屏、倍速、第 N 集。典型说法: 「用MCP打开B站播放奥特曼」「在B站看/播放」「2倍速全屏」。
禁止事项
- 禁止内置
browser_browse/browser_interact。 - 禁止反复
browser_tabs action=new;已知 URL 用browser_navigate。 - 禁止靠连点播放器控件碰运气设倍速或全屏。
- 禁止点击屏幕上的「全屏」按钮当主路径。
- 禁止对 snapshot artifact 做
read_file。 - 全屏 GPU 层导致
browser_screenshot黑屏时,禁止反复截图;改用 canvas 抓帧。
工作流
第 0 步 — 本 skill 已加载后直接做用户任务
不要先检查扩展、端口或文件哈希。
第 1 步 — 找到视频
- 用户给了 URL:
browser_navigate直达。 - 只有关键词:navigate 到
https://search.bilibili.com/all?keyword=<url-encoded>。 - 搜索快照里选一条
href以/video/BV...开头的结果,再 navigate 到完整https://www.bilibili.com/video/BV.../。优先按 URL 导航,不要点搜索卡片(点击容易让标签超时)。
合集 / 分 P /「第 N 集」:navigate 到 https://www.bilibili.com/video/BV...?p=N,
或在播放列表点对应分 P。用页面 title 或 URL 含 ?p=N 确认。
整次任务最多 browser_tabs action=new 一次。
第 2 步 — 确认正在播放
browser_snapshot 看到播放器控件即可。用 hawkeye_evaluate 读 video,不要盲点播放键:
(() => { const v = document.querySelector('video'); return v ? {paused:v.paused, cur:v.currentTime, dur:v.duration, rate:v.playbackRate} : {found:false}; })()
paused === true 时再 document.querySelector('video').play()。
过几秒再读一次 currentTime,必须递增才算在播。
第 2b 步 — 倍速(用户明确要求时)
主路径:直接设 video.playbackRate。不要点「倍速」菜单,也不要把 browser_select_option 当主路径。
该属性在进入全屏后仍然有效。
(() => { const v = document.querySelector('video'); if (v) v.playbackRate = 2; return { rate: v && v.playbackRate }; })()
返回 rate: 2 即成功。清晰度下拉才用 browser_select_option。
第 3 步 — 真实全屏
全屏需要真实用户手势(navigator.userActivation)。
主路径: 先聚焦播放器,再 browser_press_key key=f inputMode=trusted。
未传 inputMode 时 DeepSentry 会自动补 trusted。
随后 hawkeye_evaluate 校验:
(() => ({ fs: !!document.fullscreenElement, el: document.fullscreenElement && document.fullscreenElement.className, ua: (()=>{const u=navigator.userActivation;return u?{isActive:u.isActive}:null})() }))()
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 2mo ago First seen · 142 lines · 100 tokens per session scan A 0a861895447e
bilibili-play is a skill published in the GitHub repository asaotomo/DeepSentry (39 stars, last pushed 17d ago), licensed Apache-2.0. It adds 100 tokens to every session and 1,779 once invoked, about $0.0004 per session on Opus 5.5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-09-04.
bilibili-play compared with skills from other repositories
| Mod | Type | Stars | Last push | Licence | Scan | Tokens | Works with |
|---|---|---|---|---|---|---|---|
| bilibili-playasaotomo/DeepSentry | Skill | 39 | 17d ago | Apache-2.0 | A | 100 | Claude Code, Codex |
| agent-browserTencent/WeKnora | Skill | 37k | 3d ago | none | A | 83 | Claude Code, Codex |
| computer-useTheSyart/emperor-agent | Skill | 213 | yesterday | MIT | A | 99 | Claude Code, Codex |
| go-expertpersonamanagmentlayer/pcl | Skill | 44 | 11d ago | Apache-2.0 | A | 65 | Claude Code |
| web-accessitwanger/PaiCLI-Python | Skill | 113 | 11d ago | MIT | A | 32 | Claude Code, Codex |
| browser-lookuphelixml/helix | Skill | 829 | yesterday | none | A | 52 | Claude Code, Codex |
| docs-link-checkerrysh-ai/rysh-cli-code | Skill | 5 | 3d ago | Apache-2.0 | A | 15 | Claude Code, Codex |