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 Rimagination/bili-note --skill bili-notegit clone --depth 1 https://github.com/Rimagination/bili-noteWrote 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/rimagination/bili-note/bili-note)<a href="https://agentmods.dev/skills/rimagination/bili-note/bili-note"><img src="https://agentmods.dev/badge/skills/rimagination/bili-note/bili-note/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/rimagination/bili-note/bili-note"><img src="https://agentmods.dev/badge/skills/rimagination/bili-note/bili-note.svg" alt="Reviewed on agentmods" width="80" height="20"></a>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 | $0.00086 | $0.07388 |
| Opus 5.5 | $0.00034 | $0.02955 |
| Sonnet 5.5 | $0.00017 | $0.01478 |
| Haiku 4.5 | $0.00009 | $0.00739 |
Grade A, and why
bili-note scanned grade A with 1 finding 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 29d 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.
Makes network callslowCapability
Not a fault in itself. Listed so you know the mod talks to something, and to what.
curl.exe -s http://localhost:3456/targets How it starts
The opening of the file, as written. The whole thing — 391 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Bili Note
把 B站视频和图文动态变成可检索、可复用的 Markdown 知识笔记。视频优先拿字幕;字幕拿不到时再转写音频;图文优先抓正文、图片和代码块;用户要评论区时抓取评论并过滤无关讨论。
联网或登录态操作必须先使用 web-access。
网页 AI 字幕支持两条路线:Chrome 使用 web-access,Edge 使用开启远程调试的 Chromium DevTools Protocol 直连。两条路线都让已登录的 B站页面自己请求字幕接口,不读取或复制 Cookie/profile。Playwright 临时浏览器不属于受支持路线。
Bili Note 与 DyNote 共享可复用本地资源。默认共享目录是 %USERPROFILE%\.cache\rimagination-notes,Qwen3-ASR 环境默认是 %USERPROFILE%\.cache\rimagination-notes\qwen3-asr-venv。如果任一 skill 已经安装过 Qwen3-ASR,另一个 skill 必须优先复用,不要重复安装。Hugging Face、Whisper 和 faster-whisper 缓存按本机通用缓存复用。
字幕密度与视觉理解
B 站视频优先拿字幕,但字幕/转写不等于完整理解。长视频如果字幕字数明显偏少,可能说明核心信息在画面、PPT、板书、代码演示、屏幕操作或无解说片段中。
- 每次视频运行都必须先做轻量预检,再询问用户是否开启“关键帧视觉理解”。预检只读取元数据、分P信息和已有字幕文件,不下载视频;它会显示时长、字幕字数、字幕密度和画面依赖建议。收到明确的
on或off后才能进入后续流程。询问时说明:on会抽取最多 12 帧,先合成为一张 4×3 联系图,再按需要打开单帧,视觉输入会增加 tokens;off只走字幕、转写、评论和元数据路线,节省视觉输入 tokens。 - 用户选择
on时,给run_bili_note.py传--visual-review on;选择off时传--visual-review off。脚本默认--visual-review ask,直接在终端运行也会在轻量预检后停下来等待选择。 - 关键帧理解选择只对视频运行生效。图文/动态已有图片材料,按图文视觉理解流程处理。
- 归档后必须读取
metadata/note_budget.json。如果visual_dependency.needs_visual_review=true,不要把稀疏字幕写成完整学习笔记。 - 这种情况下,先采用更合适的技术:抽取代表性关键帧或截图,再做 OCR 或多模态视觉理解;必要时重新抓网页 AI 字幕或补本地自动语音识别。
- 如果当前接入的模型或工具不能看图,必须告诉用户:关键帧理解属于高级视觉能力,需要多模态模型、OCR 工具或人工查看;当前只能基于字幕/元数据生成有限笔记并标注覆盖范围。
- 最终笔记和用户回复要转述
visual_dependency.warnings,并说明是否已经补了视觉证据。
关键帧视觉理解采用低成本的两阶段读取:先看一张 4×3 联系图建立全片画面地图;联系图中出现文字、代码、图表或界面细节时,再依据工作目录的 keyframes_manifest.json 打开对应单帧。归档后对应路径是 archive_dir/metadata/keyframes_manifest.json。不要默认把 12 张完整图片全部送入模型。
选择 on 后,先查看 keyframes/contact_sheet.png,再结合 keyframes/README.md 和工作目录的 keyframes_manifest.json 定位时间点。把视觉观察写入材料包的 metadata/visual_review.md,每条关联 KF-Pxx-xx、分P、时间和观察结论;不确定的画面标记为待核实。关键帧证据可通过 indexes/关键帧索引.jsonl 和总证据索引回查。
依赖与环境检查
默认路线尽量零第三方 Python 依赖:视频元数据、公开字幕、图文正文、图片清单、评论、归档、证据索引和笔记预算都用标准库完成。网页 AI 字幕和音频转写是增强路线,不是启动门槛。
第一次使用、换机器、用户怀疑依赖不全,或准备使用网页 AI 字幕 / 音频转写兜底时,先运行:
$skill = "$env:USERPROFILE\.codex\skills\bili-note"
$py = "python"
& $py "$skill\scripts\check_environment.py"
What ships with it
15 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
- agents/openai.yaml 674 B
- assets/bili-note-logo.png 816 KB
- references/bilibili-api-notes.md 3.9 KB
- scripts/archive_bili_materials.py 54 KB runs code
- scripts/check_environment.py 18 KB runs code
- scripts/edge_cdp.py 6.3 KB runs code
- scripts/extract_bilibili_opus.py 22 KB runs code
- scripts/extract_bilibili.py 39 KB runs code
- scripts/extract_video_keyframes.py 16 KB runs code
- scripts/fetch_browser_ai_subtitles.py 13 KB runs code
- scripts/run_bili_note.py 30 KB runs code
- scripts/run_qwen_asr.py 5.9 KB runs code
- scripts/score_bili_note.py 5.4 KB runs code
- scripts/setup_qwen_asr_env.py 2.3 KB runs code
- scripts/update_note_budget_section.py 8.6 KB runs code
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.
- 29d ago Changed · +45 lines 6942882c6c7e
- 1mo ago First seen · 346 lines · 86 tokens per session scan A c4d80165a0bb
bili-note is a skill published in the GitHub repository Rimagination/bili-note (330 stars, last pushed 1mo ago), licensed MIT. It adds 86 tokens to every session and 7,388 once invoked, about $0.0003 per session on Opus 5.5. A static security scan graded it A with 1 finding (makes network calls). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other skills, from other repositories
docs-scraper
A tool-assisted method for downloading an entire technical documentation website as local Markdown files. It first checks whether the site offers ready-made Markdown or text bundles before crawling pages.
larksnap-fetch
A bridge for downloading Feishu/Lark documents or ordinary webpages into local files, with options such as Markdown, HTML, or PDF. Feishu, also called Lark, is a workplace collaboration platform.
install-by2kb
Install, configure, update, or diagnose the by2kb CLI for this Hermes plugin. Preserve existing configuration and keep Hermes-managed plugin code separate from CLI upgrades.
markdown-documents
Write or edit Markdown notes, reports and plain text; default for non-Office writing.
cv-update-review
Run a confirmation-gated CV update workflow for Word CV documents: set up a reusable profile, scan configured local activity sources plus public evidence such as PubMed through NCBI E-utilities and configured website/RSS feeds, prepare a CV update packet, ask the user to approve completed/public/accepted items, create…
rag-prep
A document-preparation skill that cleans and structures source files for retrieval systems, which find relevant passages from a document collection.