Creator · Rimagination
Last updated · Sep 6, 2026
DyNote: systematically and efficiently extract raw Douyin/DY video data and analyze videos, comments, accounts, hashtags, and short-video scenes into evidence-graded learning notes, summaries, research briefs, scripts, and knowledge-base material. Use when the user asks to 抓取/提取/
Creator · Rimagination
Last updated · Sep 6, 2026
DyNote: systematically and efficiently extract raw Douyin/DY video data and analyze videos, comments, accounts, hashtags, and short-video scenes into evidence-graded learning notes, summaries, research briefs, scripts, and knowledge-base material. Use when the user asks to 抓取/提取/
Creator · Rimagination
Last updated · Sep 6, 2026
DyNote: systematically and efficiently extract raw Douyin/DY video data and analyze videos, comments, accounts, hashtags, and short-video scenes into evidence-graded learning notes, summaries, research briefs, scripts, and knowledge-base material. Use when the user asks to 抓取/提取/
Creator · Rimagination
Last updated · Sep 6, 2026
DyNote: systematically and efficiently extract raw Douyin/DY video data and analyze videos, comments, accounts, hashtags, and short-video scenes into evidence-graded learning notes, summaries, research briefs, scripts, and knowledge-base material. Use when the user asks to 抓取/提取/
Do not auto-install
Install targets
Codex install prompt
Install the "dy-note" agent skill from https://github.com/Rimagination/dy-note/blob/main/SKILL.md. Read its SKILL.md or equivalent instructions first, install only the files needed for this workspace, and summarize any required setup before using it. Skill purpose: DyNote: systematically and efficiently extract raw Douyin/DY video data and analyze videos, comments, accounts, hashtags, and short-video scenes into evidence-graded learning notes, summaries, research briefs, scripts, and knowledge-base material. Use when the user asks to 抓取/提取/整理 抖音视频字幕、视频文案、ASR 转写、Qwen3-ASR 中文转写、原始材料归档、学习笔记、analysis plan、note budget、避免返工、复用已有素材、评论洞察、账号分析、赛道/话题研究、竞品拆解、电商/本地生活视频分析、事实核查、自动搜索素材, save Douyin content as Markdown/TXT, or use subtitle/local ASR as the factual spine with logged-in Douyin Web built-in AI / Doubao fallback as visual or quick-reading supplements. After the install attempt, report the result to https://www.openagentskill.com/api/agent/outcome with POST JSON {"event_id":"install_<unique-id>","skill_slug":"rimagination-dy-note","task":"Install dy-note","agent":"codex","outcome":"success","install_used":true}. Replace event_id with a unique value and outcome with success or failed. Report success only after the skill is installed and a minimal verification passes.Supply asset profile
Deep research, source comparison, literature review, RAG, knowledge search, and reports.
Scenario
Research agents
I need my agent to research a topic, compare sources, and produce a concise report.
Agent fit
Claude Code + OpenAI Agents + Browser agents
Codex, Claude Code, Cursor, CLI, or custom agents.
Install
Ready
npx skills add Rimagination/dy-note --skill dy-note
Maintenance
active
2mo since push
Risk
Needs review
Dependency or permission surface needs review
GitHub quality
159
63/100 Quality · 63/100 Trust
Coverage tags
Review notes
Dependency or permission surface needs review · Permission surface may require sandboxing
Agent adoption scorecard
These scores combine public repository metadata, OpenAgentSkill review signals, maintenance freshness, and install readiness. They are a shortlist signal, not a replacement for human review.
Quality
PromisingUseful candidate, but compare it with alternatives before adopting.
Trust
Do not auto-installTrust Score v5 found insufficient evidence for agent installation. Treat this as discovery material, not an executable recommendation.
Audit
Needs reviewA machine-readable review of install readiness, security metadata, maintenance, and adoption risk.
OpenAgentSkill Trust Score v5
Choose a stronger alternative or inspect the source manually before any install attempt.
Stars
159 GitHub stars
Repo activity
159 stars, 24 forks
Maintenance
2mo since push
License
MIT
Install
npx skills add Rimagination/dy-note --skill dy-note
Install safety
Agent-readable metadata
Use this block or the embedded JSON to decide whether an agent should install this skill, choose an alternative, or ask for human review first.
Suited tasks
Suited agents
Install decision
Trust and risk
Outcome loop
Install command
npx skills add Rimagination/dy-note --skill dy-noteDo not use when
Alternative
1.9K Stars
npx skills add yanliudesign/mono-color-skill --skill mono-color
Alternative
61.0K Stars
npx skills add mvanhorn/last30days-skill -g
Alternative
38.4K Stars
npx skills add Imbad0202/academic-research-skills
Alternative
256.3K Stars
npx skills add mattpocock/skills --skill grill-me
Agent safety v2
This skill should not be selected by an agent without explicit human security review.
Do not auto-install. Inspect the source, dependencies, and permission surface first.
high
Skill metadata references terminal, CLI, shell, subprocess, or command execution workflows.
medium
Skill may drive a browser or interact with web pages.
medium
Skill likely fetches remote pages, APIs, repositories, or external services.
medium
Skill may read or write project files, documents, generated artifacts, or local workspace state.
Agent resolve plan
The Resolve API returns the selected skill, alternatives, safety policy, audit notes, install target, and copy-paste prompt an agent can follow without scraping this page.
Open JSON
/api/agent/resolve?task=Use%20dy-note%20for%20an%20agent%20workflow&agent=codex&max_risk=medium
Resolve text
/api/agent/resolve?task=Use%20dy-note%20for%20an%20agent%20workflow&agent=codex&max_risk=medium&format=text
Install handoff
/api/skills/rimagination-dy-note/install
Agent should check
Copy prompt
Task: Use dy-note in this workspace.
Resolve first: https://www.openagentskill.com/api/agent/resolve?task=Use%20dy-note%20for%20an%20agent%20workflow&agent=codex&max_risk=medium
Review install handoff: https://www.openagentskill.com/api/skills/rimagination-dy-note/install
Install command: npx skills add Rimagination/dy-note --skill dy-note
Before running it, summarize audit warnings, required permissions, and the fallback skill if install is risky.Agent handoff
Use the public install endpoint to fetch the command, safety checklist, target prompts, and canonical links for this skill.
Install handoff
/api/skills/rimagination-dy-note/install
LLM text format
/api/skills/rimagination-dy-note/install?format=text
Find alternatives
/api/skills/search?q=dy-note&limit=3
Agent prompt
Use dy-note for this task. Review https://www.openagentskill.com/api/skills/rimagination-dy-note/install, then install with: npx skills add Rimagination/dy-note --skill dy-noteRegistry metadata
This page exposes the same decision, trust, audit, use-case, and install signals through the Registry API, so agents can rank this skill without scraping the UI.
Manifest
/api/registry/manifest/rimagination-dy-note
LLM text
/api/registry/manifest/rimagination-dy-note?format=text
Install alias
/api/registry/install/rimagination-dy-note
Recommend
/api/registry/recommend?task=Use%20dy-note%20in%20an%20agent%20workflow&limit=3
Agent fit
Research agents
Use-case tags
Platforms
Claude Code, OpenAI Agents, Browser agents
Audit report
A machine-readable review of install readiness, security metadata, maintenance, and adoption risk.
Agent decision cockpit
Prototype with this skill first; keep a fallback candidate ready.
Role in stack
Fallback candidate
Primary fit
Research agents
Trust label
Prototype first
Install path
Command ready
Use when
Evidence
review first
Implementation path
Trust profile
Trust Score v5 found insufficient evidence for agent installation. Treat this as discovery material, not an executable recommendation.
GitHub adoption
INFO159 GitHub stars
Stars/forks activity
CHECK159 stars, 24 forks; issue activity unavailable in current metadata
Recent maintenance
PASS2mo since push
License clarity
PASSMIT
Good signals
Review before install
Recommended action
Choose a stronger alternative or inspect the source manually before any install attempt.
Quality profile
Useful candidate, but compare it with alternatives before adopting.
Workflow fit
Investigate faster
I need my agent to research a topic, compare sources, and produce a concise report.
Publish consistently
I need my agent to turn research and product updates into useful content drafts.
Process rich media
I need my agent to process images, video, or audio and extract useful information.
Workflow fit
Find, compare, and synthesize
A workflow for agents that gather sources, compare claims, summarize long material, and draft useful research briefs.
Turn skills into distribution
A workflow for turning newly indexed skills into SEO briefs, social drafts, comparison pages, and reusable publishing workflows.
Scrape, clean, and reuse web data
A practical workflow for agents that crawl public pages, extract clean content, normalize data, and hand it to downstream research or RAG workflows.
Alternative shortlist
Similar skills that may fit this task.
Generate original one-ink or controlled two-ink editorial images from any theme, sentence, article idea, object, or reference photo. Always use this skill when the user asks for 单色海报、双色印刷、单色调视觉、蓝色/绿色孔版印刷、risograph、网点照片、复古或当代编辑排版、zine poster, monochrome editorial poster, duotone print, or asks to use the mono-color style. It uses an adaptive white, gray, or pale-beige substrate, no more than two printing inks, active negative space, terse human language, and strong serif/grotesk/mono typography without making retro styling the default or copying a source composition, wording, logo, or artwork. Produce both the final generation prompt and the generated raster image unless the user explicitly asks for prompt only.
Research the last 30 days across Reddit, X, YouTube, Hacker News, Polymarket, GitHub, and the web, then synthesize a grounded brief for an AI agent.
Academic Research Skills for Claude Code: research → write → review → revise → finalize
A relentless interview to sharpen a plan or design.
--- name: dy-note description: "DyNote: systematically and efficiently extract raw Douyin/DY video data and analyze videos, comments, accounts, hashtags, and short-video scenes into evidence-graded learning notes, summaries, research briefs, scripts, and knowledge-base material. Use when the user asks to 抓取/提取/整理 抖音视频字幕、视频文案、ASR 转写、Qwen3-ASR 中文转写、原始材料归档、学习笔记、analysis plan、note budget、避免返工、复用已有素材、评论洞察、账号分析、赛道/话题研究、竞品拆解、电商/本地生活视频分析、事实核查、自动搜索素材, save Douyin content as Markdown/TXT, or use subtitle/local ASR as the factual spine with logged-in Douyin Web built-in AI / Doubao fallback as visual or quick-reading supplements." ---
# DyNote
DyNote 是面向 Codex 这类 Agent 的抖音学习工具,不是一次性摘要器。核心原则是“数据资产先行,学习笔记后置”:先把字幕/转写、评论、元数据和 AI 快读沉淀为可复用资产,再按用户需求生成可追溯的学习笔记、总结和写作材料。默认目标不是字幕工程文件,而是先落一份原始数据包:`douyin_ai_brief.md`、`douyin_ai_brief.json`、`transcript.cleaned.md`、`transcript.txt`、`segments.json`、`metadata.json`、`note_budget.json`,并用 `assets/` 归档可复用资产。默认把独立字幕轨或本地自动语音识别转写当作事实主干;当转写密度低、任务需要画面理解或用户只要快速筛选时,再用已登录抖音网页版的“问AI / 识别画面”补充,豆包只作为抖音 AI 不可用时的备用快读或待核验假设。
联网或登录态操作必须先使用 `web-access`。不要读取、复制或打印 Cookie、msToken、a_bogus、x-secsdk-web-signature、临时签名视频 URL 等敏感参数;脚本只让已授权 Chrome 页面自己加载内容。
DyNote 与 Bili Note 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。
## 浏览器与登录态硬规则
- 抖音内置 AI 和豆包备用路线都只使用 `web-access` 连接到用户当前可用的 Chrome。不要启动无登录态 Playwright 浏览器,不要用静态 curl 抓登录页,不要导出或保存 `storageState`、Cookie、localStorage 或 token。 - 抖音内置 AI 路线需要当前 Chrome 已登录抖音网页版。它主要用于低转写密度、画面文字、镜头/场景或快速筛选;打开视频后使用页面右侧 `问AI`,必要时点击可见的 `识别画面` 把当前帧加入问答上下文。 - 如果抖音页面没有 `问AI` / `识别画面` 或未生成 `章节要点`,记录为 `weak` 或 `blocked`,先确保字幕/本地自动语音识别事实主干可用,再考虑豆包备用、关键帧或 OCR。 - 在向豆包发送内容前,必须确认 `https://www.doubao.com/chat/` 在当前 Chrome 中已登录且有可见聊天输入框、侧边栏/新对话等用户态界面。 - 如果未检测到豆包登录态,停止并返回 `blocked: doubao-login-required`,提示用户先在同一个 Chrome 登录豆包。不要静默降级到其他浏览器。 - 豆包备用快速解读优先使用用户复制的完整抖音分享文本,不要只喂最终 `douyin.com/video/...`,因为完整分享文案更容易触发豆包的搜索/参考资料式视频概述。 - 豆包输出需要做证据分级:`search-derived` 是检索式概述,`visual-claimed` 是声称包含画面/镜头细节,`blocked` 是豆包无法访问视频画面,`weak` 是信息不足。不要把检索式概述说成逐帧视觉解析,也不要让豆包替代完整字幕或本地转写。
## 抖音字幕现实与默认路线
抖音和 B 站不同:很多抖音视频没有可直接抓取的独立字幕文件。先按用户任务分流,不要固定把所有视频都下载转写。
- 用户只是问“这个视频讲什么”、想做选题筛选、草稿或快速理解时,可以走已登录抖音网页版内置 AI 快读;但输出必须标注为快读/视觉假设,不写成完整字幕提取。已有 `douyin_ai_brief.json` 时先复用;如果不可用,再用豆包 `fast` 备用。 - 用户要学习笔记、可靠原文、逐句内容、引用、脚本拆解、事实核查或可发布材料时,先找已有 SRT/VTT/TXT;没有可用字幕轨或转写时,主要依赖本地自动语音识别。中文或未指定语言优先共享 Qwen3-ASR,明确外语视频再用 Whisper 系后端。 - 用户要镜头、画面文字、贴纸文字、操作步骤、商品/价格/场景细节,或 `note_budget.json` 显示转写密度低时,不能只靠音频转写;优先补抖音 `问AI / 识别画面`、关键帧、截图或 OCR。抖音 AI 仍不能提取完整字幕;豆包 `evidence` 只作为抖音 AI 不行时的备用视觉假设。
抖音字幕常见两种形态:
- 独立字幕轨:创作者使用平台字幕功能或上传 SRT 后,网页播放器可能叠加渲染 VTT 字幕。若能抓到这类轨道,可作为逐句文本材料。 - 画面内嵌文字:字幕、贴纸或手动排版文字已经焊在画面里,没有独立文件。音频转写只能识别人声,不能读取这类画面文字,必须补视觉证据。
如果视频较长,但 `note_budget.json` 中 `visual_dependency.risk` 为 `medium` 或 `high`,必须提醒用户:转写文本过少,完整理解可能依赖画面,不能把稀疏本地自动语音识别结果写成完整笔记。抖音问 AI、豆包和关键帧结果必须作为补充证据审计,不能替代原文主干。
## 场景模式路由
先判断用户真正要完成的任务,再决定证据深度和工具路线:
- `single-video-note`:默认模式。单条视频/分享文本 -> 字幕/本地自动语音识别做事实主干;转写稀疏或任务需要画面时补抖音问 AI、关键帧/OCR;抖音 AI 不可用时豆包才作为备用假设,输出可读笔记。 - `comment-insight`:用户关心评论、痛点、需求、FAQ、反对意见或爆点反馈时,加载 `douyin-comments`。默认只抓前 100 条主评论及这些主评论的楼中楼,输出 `_sample.json/csv`,并明确提示这不是全部评论。用户明确要完整评论资产、复核全部可见评论或样本不足时,再用 `--full` 做全量抓取。抓到的 JSON/CSV 必须归档到 `assets/comments/`,再输出用户洞察,而不是只把评论写进一次性总结。 - `account-analysis`:账号主页或多条视频 -> 定位、内容支柱、钩子模板、系列化栏目、发布节奏和可复用选题。 - `topic-research`:话题、关键词、赛道、竞品或“自动搜索” -> 先低成本收集标题/简介/话题/样本链接,再按需要升级到 ASR、评论和关键帧。 - `script-mining`:拆脚本、镜头、叙事节奏、开头钩子、转场、结尾 CTA;脚本文案以字幕/本地转写为主,抖音内置 AI 或备用豆包只给画面假设,重要结论要用转写/抽帧校验。 - `commerce-analysis`:带货、本地生活、探店、课程或服务视频 -> 卖点、信任证据、价格/优惠、CTA、转化阻力和评论需求。 - `fact-check`:涉及医学、法律、投资、新闻或强事实判断时,区分视频原文、抖音内置 AI/豆包概述和外部来源;高风险结论必须联网核验并标注来源。 - `knowledge-archive`:用户要沉淀资料库、Obsidian、RAG 或写作素材时,保留来源 URL、作者、时间、证据等级、关键词和后续可检索标签。
成本分层默认从轻到重:
- `quick-pass`:分享文本、页面元数据、抖音内置 AI,必要时备用豆包 `fast`;适合秒级判断、选题筛选和草稿,但必须标注不是完整字幕/全文证据。 - `evidence-pass`:独立字幕轨或 ASR 全文、关键帧/OCR、评论样本;适合要引用、拆解或发布的内容。 - `research-pass`:批量视频、账号/话题搜索、竞品对比和评论聚类;范围大时先给样本计划和 token/时间风险。
## 系统化分析协议
默认按“问题 -> 取证 -> 分析 -> 审计”推进,不要把工具输出直接等同于结论:
1. `research-question`:写清要回答的问题、分析单位和场景模式。复杂任务先生成 `analysis_plan.json`。 2. `sampling-plan`:账号、话题、评论或竞品任务必须说明样本怎么选、样本量是多少、为什么足够或不足。 3. `evidence-ladder`:把证据分为用户输入、页面元数据、独立字幕轨/本地自动语音识别转写、抖音内置 AI、备用豆包快读、评论、关键帧/OCR、外部来源。字幕/转写是事实主干;抖音 AI 和豆包是快读或视觉补充。结论必须标注依赖哪一层。 4. `synthesis-gate`:合成前先读取 `assets/asset_manifest.json` 或确认同等原始材料,检查证据等级、覆盖范围和反例/不确定性;缺证据时先写范围限制,不要补故事。 5. `audit-trail`:最终笔记或研究简报保留来源 URL、采集时间、输出文件、样本范围、`note_budget.json` 和无法验证的点。
## 高效执行与复用策略
- 先检查已有产物,再决定下一步。已有 `douyin_ai_brief.json` 时,不要重复问抖音内置 AI;已有 `doubao_brief.json` 时,不要重复问豆包;已有 `transcript.txt`、`segments.json`、`metadata.json` 时,不要重跑 ASR;已有 `note_budget.json` 且未过期时,不要重算预算。 - `analysis_plan.json` 只在复杂任务或目标变化时创建;已有计划默认复用。目标、来源、模式或证据等级变化时才用 `--force` 重建。 - 先走最便宜的 `quick-pass`,只有当研究问题无法回答、证据等级不足、或用户要可发布笔记/事实核查时,才升级到 `evidence-pass` 或 `research-pass`。评论区任务的 `quick-pass` 是前 100 条主评论及对应楼中楼样本;全量可见评论属于更重的资产补齐步骤。 - 不要为了“完整流程”固定执行所有步骤。单条视频如果已有高质量转写,可直接预算和写笔记;评论洞察如果只问观众反馈,可以先抓 100 条评论样本,不必先全量 ASR,也不必默认抓完整评论区。 - 重新运行昂贵步骤前必须说明触发条件:输入变了、旧文件缺失/损坏/过期、证据等级不足,或用户明确要求更高质量。
## 原始材料与学习笔记默认策略
- 默认先建立数据资产,再写学习笔记。不要只把抖音内置 AI、豆包概述或未经审计的本地自动语音识别文本直接当最终笔记。 - `douyin_ai_brief.json`、`doubao_brief.json`、`transcript.txt`、`segments.json`、`metadata.json`、评论 JSON/CSV 和关键帧截图都属于原始数据;最终学习笔记必须能回到这些材料解释来源。 - 默认把字幕/转写和完整评论整理成资产包:`assets/transcripts/` 保存字幕、转写和片段;`assets/comments/` 保存完整评论 JSON/CSV、JSONL 明细和可读 Markdown;`assets/asset_manifest.json` 是后续再分析的入口。 - `assets/asset_manifest.json` 是事实入口;`learning_note.md` 是从资产生成的一种学习视图,不是资产本身。用户换问题、换场景或要求复核时,优先复用资产重新组织笔记,不要重跑或覆盖原始材料。 - 写笔记前先确认用户需求:内容学习、脚本复盘、评论洞察、事实核查、写作素材或知识库归档。再从资产中选择证据和结构,不要先脑补结论再找材料。 - 每次生成转写材料后读取 `note_budget.json`。它根据视频时长、转写字数、片段/证据块、评论数和互动质量给出推荐笔记长度。 - 长视频、长转写、高评论或高互动视频要写更长、更结构化的笔记;短视频或低信息密度视频避免过度扩写。 - 互动质量是“值得多写”的辅助信号,不替代证据。扩写必须来自视频原文、评论样本、关键帧、外部核验,或已明确降级为视觉假设的抖音内置 AI/豆包快读。 - 学习型笔记要让读者像学完一个短课题:获得概念、判断标准、方法步骤、适用边界、坑点和可迁移用法。不要只写“视频讲了什么”。
## 默认流程
1. 读清用户要的是哪种场景模式:单条视频笔记、评论洞察、账号分析、话题研究、脚本拆解、电商分析、事实核查,还是知识库归档。 2. 首次使用、换机器、或准备跑完整抖音链接流程时,先检查环境:
```powershell $skill = "$env:USERPROFILE\.codex\skills\dy-note" $py = "python" & $py "$skill\scripts\check_environment.py" ```
3. 如果有输出目录,先运行或心中执行 `inspect_workflow_state.py`,复用已有计划、抖音内置 AI brief、豆包 brief、转写、评论、预算和评分。 4. 评论、账号、话题、竞品、事实核查或批量研究任务先用 `create_analysis_plan.py` 生成 `analysis_plan.json`;已有计划且目标没变时不要重建。单条视频任务也要在脑中执行同样的证据闸门。 5. 如果已经有 SRT、Whisper JSON 或 TXT,优先走本地整理路线,避免重复下载和转写;脚本会生成 `note_budget.json`。 6. 如果用户要求可靠全文、学习笔记、逐句内容、引用、脚本拆解或事实核查,优先使用 `extract_douyin_text.py` 取得字幕轨或本地自动语音识别转写。默认 `--asr-backend auto`:中文或未指定语言优先共享 Qwen3-ASR,明确外语视频优先 Whisper。默认复用已有输出;需要重跑时加 `--force`。 7. 如果用户只是快速理解、选题筛选或先拿草稿,且当前 Chrome 已登录抖音网页版,可以跑 `douyin_web_ai_brief.py`;已有可用 `douyin_ai_brief.json` 时先读旧结果。抖音内置 AI 不可用、弱,或 `note_budget.json` 显示低转写密度时,再用 `doubao_video_brief.py` 的 `fast/evidence` 模式备用,但必须标注为假设。 8. 评论、账号、话题、竞品或批量研究任务先做 `quick-pass` 样本,不要直接对大量视频逐条 ASR;样本结论不足时再升级到 `evidence-pass` 或 `research-pass`。大评论区默认用 `fetch_douyin_comments.py` 抓前 100 条主评论及对应楼中楼;需要完整可见评论时再显式加 `--full`,不要黑盒等待一轮全量抓取。 9. 每次得到转写、字幕、评论 JSON/CSV、抖音 AI brief 或豆包 brief 后,运行 `archive_dy_note_assets.py` 生成或更新 `assets/`。评论区任务必须保留评论样本或完整评论资产,不能只输出评论摘要。 10. 写学习笔记前,先打开 `assets/asset_manifest.json`,再按用户需求读取 `transcript.cleaned.md`、`segments.json`、完整评论、`douyin_ai_brief.md`、`doubao_brief.md` 或 `analysis_plan.json`,检查证据是否足以回答研究问题。再打开 `note_budget.json`,按推荐长度和 `writing_guidance` 写学习型笔记;如果 `visual_dependency.needs_visual_review=true`,必须在笔记和回复中提醒画面证据不足,或先补抖音 `问AI / 识别画面`、关键帧/OCR。对外部 AI 回答逐项做证据审计,不要把它们的扩写混成视频原文。 11. 如果识别出片名、人名、地名明显错,优先在最终说明里标注可疑词;用户要求校对时再做替换、二次 ASR 或补关键帧。 12. 用户明确要 SRT/VTT/时间轴时,才单独交付字幕文件;无论是否单独交付,都要在 `assets/transcripts/` 里保留可复用文本资产。
## 常用命令
### 0. 检查已有工作状态
已有输出目录时先检查状态,决定下一步,不要盲目重跑:
```powershell & $py "$skill\scripts\inspect_workflow_state.py" ` --out-dir ".\dy_note_output" ` --mode "single-video-note" ```
重点看:
- `reusable_artifacts`:可以直接复用的产物。 - `recommended_next_steps`:真正缺的下一步。 - `avoid_rework`:明确不要重复做的昂贵步骤。 - `stale.note_budget` / `stale.note_score`:预算或评分是否因新材料而过期。
### 1. 创建系统化分析计划
复杂任务先生成计划,再采集数据。单条视频可省略显式文件,但账号/话题/评论/竞品/事实核查任务建议保留:
```powershell & $py "$skill\scripts\create_analysis_plan.py" ` --mode "topic-research" ` --tier "quick-pass" ` --objective "分析这个赛道里什么视频形式值得复用" ` --source "铁板牛排 炸土豆饼 野外烹饪" ` --out-dir ".\dy_note_research" ```
如果 `analysis_plan.json` 已存在,脚本默认复用旧计划;只有目标、来源、模式或证据等级改变时才加 `--force` 覆盖。
常用模式:`single-video-note`、`comment-insight`、`account-analysis`、`topic-research`、`script-mining`、`commerce-analysis`、`fact-check`、`knowledge-archive`。
### 2. 从已有 SRT 生成干净文本
```powershell & $py "$skill\scripts\extract_douyin_text.py" ` --from-srt "D:\微信推送\douyin_subtitles_7647145112421633320\7647145112421633320_16k.srt" ` --metadata-json "D:\微信推送\douyin_subtitles_7647145112421633320\official_detail_summary.json" ` --out-dir "D:\微信推送\dy_note_7647145112421633320" ```
### 3. 从 Whisper JSON 或 TXT 生成干净文本
```powershell & $py "$skill\scripts\extract_douyin_text.py" ` --from-whisper-json ".\audio_16k.json" ` --out-dir ".\dy_note_output"
& $py "$skill\scripts\extract_douyin_text.py" ` --from-txt ".\raw_transcript.txt" ` --source-url "https://www.douyin.com/video/..." ` --out-dir ".\dy_note_output" ```
### 4. 安装或检查共享 Qwen3-ASR 本地环境
Qwen3-ASR 是中文视频优先使用的本地自动语音识别后端。首次使用时安装到共享 venv,复用现有 CUDA Torch;DyNote 和 Bili Note 共用同一套环境:
```powershell & $py "$skill\scripts\setup_qwen_asr_env.py" & $py "$skill\scripts\check_environment.py" ```
看到 `routes.qwen3_asr=OK` 后再使用 Qwen 后端。本机默认 venv 路径是:
```text %USERPROFILE%\.cache\rimagination-notes\qwen3-asr-venv ```
为兼容早期原型,脚本仍会探测旧路径 `%USERPROFILE%\.cache\dy-note\qwen3-asr-venv` 和 `%USERPROFILE%\.cache\douyin-note\qwen3-asr-venv`。
### 5. 从抖音链接完整提取
先按 `web-access` 要求启动并检查 CDP proxy,再运行:
```powershell & $py "$skill\scripts\extract_douyin_text.py" ` "https://v.douyin.com/xxxxxxx/" ` --out-dir "D:\微信推送\dy_note_output" ` --asr-model medium ` --language Chinese ```
如果输出目录已有 `transcript.txt`、`segments.json` 和 `metadata.json`,脚本默认复用并跳过浏览器、下载和 ASR。确实要重跑时加:
```powershell --force ```
中文长视频默认优先用 Qwen3-ASR-0.6B。8GB 显存建议保留默认 60 秒分段,避免整段长音频 OOM:
```powershell & $py "$skill\scripts\extract_douyin_text.py" ` "https://v.douyin.com/xxxxxxx/" ` --out-dir "D:\微信推送\dy_note_output_qwen" ` --asr-backend qwen3-asr ` --qwen-model "Qwen/Qwen3-ASR-0.6B" ` --qwen-chunk-seconds 60 ` --language Chinese ```
脚本会输出:
- `transcript.cleaned.md`:适合阅读和继续写笔记的 Markdown。 - `transcript.txt`:纯文本正文,适合喂给总结、RAG 或写作流程。 - `segments.json`:按原始字幕/ASR 片段保留的结构化文本。 - `metadata.json`:来源、作者、作品 ID、片段数、生成时间和输出清单。 - `note_budget.json`:按时长、转写字数、片段数、评论量和互动质量生成的推荐学习笔记长度,以及转写过稀时的画面依赖提示。 - `page_metadata.json`、视频、音频、Whisper SRT 或 Qwen JSON:完整流程产生的中间材料,供排错和回查使用。
### 6. 从已有音频使用 Qwen 转写
```powershell & $py "$skill\scripts\extract_douyin_text.py" ` --from-audio "D:\微信推送\video_16k.wav" ` --asr-backend qwen3-asr ` --qwen-chunk-seconds 60 ` --metadata-json ".\metadata.json" ` --out-dir ".\dy_note_qwen" ```
### 7. 复用已打开的 web-access target
如果已经用 `web-access` 打开抖音页面并拿到 target id:
```powershell & $py "$skill\scripts\extract_douyin_text.py" ` "https://www.douyin.com/video/7647145112421633320" ` --target "CDP_TARGET_ID" ` --keep-tab ` --out-dir ".\dy_note_output" ```
不要关闭用户已有 tab;只有脚本自己新建的 tab 可以自动关闭。
### 8. 用抖音内置 AI 快速解读视频
先按 `web-access` 要求启动并检查 CDP proxy。脚本会使用当前 Chrome 打开抖音视频页,优先读取页面 `问AI` 生成的 `章节要点` 和时间线;如果传入的是 `jingxuan?modal_id=...`,会自动归一到更稳定的 `/video/<id>` 页面:
```powershell & $py "$skill\scripts\douyin_web_ai_brief.py" ` "https://www.douyin.com/jingxuan?modal_id=7655645985318085322" ` --out-dir ".\dy_note_douyin_ai_7655645985318085322" ```
如果任务依赖当前画面内容,可尝试把暂停帧加入抖音 AI 输入框:
```powershell & $py "$skill\scripts\douyin_web_ai_brief.py" ` "https://www.douyin.com/jingxuan?modal_id=7655645985318085322" ` --identify-frame ` --out-dir ".\dy_note_douyin_ai_frame" ```
脚本会输出:
- `douyin_ai_brief.md`:抖音内置 AI 的章节要点、时间线、识别画面状态和局限。 - `douyin_ai_brief.json`:来源 URL、归一
Source provenance
Decision snapshot
recent repository activity
Audit
Install and adoption review
Agent-proven evidence
Outcome reports after resolve, review, install, and one narrow run.
No agent outcome data yet. The first agent run can report success, setup needs, risk blocks, failure, or not-relevant through /api/agent/outcome.
Install
Free and open source. Review the report before installing into production agents.
Growth loop
Scenario-led draft for dy-note, ready for a manual X post.
dy-note: DyNote: systematically and efficiently extract raw Douyin/DY video data and analyze videos, c... 159 stars https://www.openagentskill.com/skills/rimagination-dy-note?ref=x
Listing + install path for dy-note: https://www.openagentskill.com/skills/rimagination-dy-note?ref=x Install: npx skills add Rimagination/dy-note --skill dy-note
Listing source
This listing was indexed from public sources and is not marked official until a maintainer claim is approved.
Attribution links to the public repository or creator profile. Creators can claim the listing to update ownership signals.
Claim this skillOwner claim
This Registry indexed listing is attributed to Rimagination but is not marked official yet. Claim it to add a verified owner signal and make future launch, install, and audit updates easier to trust.
Creator backlink kit
Show the canonical listing, current trust and audit signals, and real Agent-Proven evidence where developers evaluate the repository.
[](https://www.openagentskill.com/skills/rimagination-dy-note?ref=github&utm_source=github&utm_medium=referral&utm_campaign=creator_badge)
[](https://www.openagentskill.com/skills/rimagination-dy-note?ref=github&utm_source=github&utm_medium=referral&utm_campaign=creator_badge)
[](https://www.openagentskill.com/skills/rimagination-dy-note/audit)
[](https://www.openagentskill.com/skills/rimagination-dy-note?ref=github&utm_source=github&utm_medium=referral&utm_campaign=creator_badge)Rimagination
@rimagination
Share whether this skill looks useful for your agent workflow. Aggregated feedback improves rankings over time.
Do not auto-install
mono-color
Generate original one-ink or controlled two-ink editorial images from any theme, sentence, article idea, object, or reference photo. Always use this skill when the user asks for 单色海报、双色印刷、单色调视觉、蓝色/绿色孔版印刷、risograph、网点照片、复古或当代编辑排版、zine poster, monochrome editorial poster, duotone print, or asks to use the mono-color style. It uses an adaptive white, gray, or pale-beige substrate, no more than two printing inks, active negative space, terse human language, and strong serif/grotesk/mono typography without making retro styling the default or copying a source composition, wording, logo, or artwork. Produce both the final generation prompt and the generated raster image unless the user explicitly asks for prompt only.
1.9K StarsLast30days Skill
Research the last 30 days across Reddit, X, YouTube, Hacker News, Polymarket, GitHub, and the web, then synthesize a grounded brief for an AI agent.
61.0K StarsAcademic Research Skills
Academic Research Skills for Claude Code: research → write → review → revise → finalize
38.4K Starsgrill-me
A relentless interview to sharpen a plan or design.
256.3K StarsDo not auto-install
Install targets
Codex install prompt
Install the "dy-note" agent skill from https://github.com/Rimagination/dy-note/blob/main/SKILL.md. Read its SKILL.md or equivalent instructions first, install only the files needed for this workspace, and summarize any required setup before using it. Skill purpose: DyNote: systematically and efficiently extract raw Douyin/DY video data and analyze videos, comments, accounts, hashtags, and short-video scenes into evidence-graded learning notes, summaries, research briefs, scripts, and knowledge-base material. Use when the user asks to 抓取/提取/整理 抖音视频字幕、视频文案、ASR 转写、Qwen3-ASR 中文转写、原始材料归档、学习笔记、analysis plan、note budget、避免返工、复用已有素材、评论洞察、账号分析、赛道/话题研究、竞品拆解、电商/本地生活视频分析、事实核查、自动搜索素材, save Douyin content as Markdown/TXT, or use subtitle/local ASR as the factual spine with logged-in Douyin Web built-in AI / Doubao fallback as visual or quick-reading supplements. After the install attempt, report the result to https://www.openagentskill.com/api/agent/outcome with POST JSON {"event_id":"install_<unique-id>","skill_slug":"rimagination-dy-note","task":"Install dy-note","agent":"codex","outcome":"success","install_used":true}. Replace event_id with a unique value and outcome with success or failed. Report success only after the skill is installed and a minimal verification passes.Supply asset profile
Deep research, source comparison, literature review, RAG, knowledge search, and reports.
Scenario
Research agents
I need my agent to research a topic, compare sources, and produce a concise report.
Agent fit
Claude Code + OpenAI Agents + Browser agents
Codex, Claude Code, Cursor, CLI, or custom agents.
Install
Ready
npx skills add Rimagination/dy-note --skill dy-note
Maintenance
active
2mo since push
Risk
Needs review
Dependency or permission surface needs review
GitHub quality
159
63/100 Quality · 63/100 Trust
Coverage tags
Review notes
Dependency or permission surface needs review · Permission surface may require sandboxing
Agent adoption scorecard
These scores combine public repository metadata, OpenAgentSkill review signals, maintenance freshness, and install readiness. They are a shortlist signal, not a replacement for human review.
Quality
PromisingUseful candidate, but compare it with alternatives before adopting.
Trust
Do not auto-installTrust Score v5 found insufficient evidence for agent installation. Treat this as discovery material, not an executable recommendation.
Audit
Needs reviewA machine-readable review of install readiness, security metadata, maintenance, and adoption risk.
OpenAgentSkill Trust Score v5
Choose a stronger alternative or inspect the source manually before any install attempt.
Stars
159 GitHub stars
Repo activity
159 stars, 24 forks
Maintenance
2mo since push
License
MIT
Install
npx skills add Rimagination/dy-note --skill dy-note
Install safety
Agent-readable metadata
Use this block or the embedded JSON to decide whether an agent should install this skill, choose an alternative, or ask for human review first.
Suited tasks
Suited agents
Install decision
Trust and risk
Outcome loop
Install command
npx skills add Rimagination/dy-note --skill dy-noteDo not use when
Alternative
1.9K Stars
npx skills add yanliudesign/mono-color-skill --skill mono-color
Alternative
61.0K Stars
npx skills add mvanhorn/last30days-skill -g
Alternative
38.4K Stars
npx skills add Imbad0202/academic-research-skills
Alternative
256.3K Stars
npx skills add mattpocock/skills --skill grill-me
Agent safety v2
This skill should not be selected by an agent without explicit human security review.
Do not auto-install. Inspect the source, dependencies, and permission surface first.
high
Skill metadata references terminal, CLI, shell, subprocess, or command execution workflows.
medium
Skill may drive a browser or interact with web pages.
medium
Skill likely fetches remote pages, APIs, repositories, or external services.
medium
Skill may read or write project files, documents, generated artifacts, or local workspace state.
Agent resolve plan
The Resolve API returns the selected skill, alternatives, safety policy, audit notes, install target, and copy-paste prompt an agent can follow without scraping this page.
Open JSON
/api/agent/resolve?task=Use%20dy-note%20for%20an%20agent%20workflow&agent=codex&max_risk=medium
Resolve text
/api/agent/resolve?task=Use%20dy-note%20for%20an%20agent%20workflow&agent=codex&max_risk=medium&format=text
Install handoff
/api/skills/rimagination-dy-note/install
Agent should check
Copy prompt
Task: Use dy-note in this workspace.
Resolve first: https://www.openagentskill.com/api/agent/resolve?task=Use%20dy-note%20for%20an%20agent%20workflow&agent=codex&max_risk=medium
Review install handoff: https://www.openagentskill.com/api/skills/rimagination-dy-note/install
Install command: npx skills add Rimagination/dy-note --skill dy-note
Before running it, summarize audit warnings, required permissions, and the fallback skill if install is risky.Agent handoff
Use the public install endpoint to fetch the command, safety checklist, target prompts, and canonical links for this skill.
Install handoff
/api/skills/rimagination-dy-note/install
LLM text format
/api/skills/rimagination-dy-note/install?format=text
Find alternatives
/api/skills/search?q=dy-note&limit=3
Agent prompt
Use dy-note for this task. Review https://www.openagentskill.com/api/skills/rimagination-dy-note/install, then install with: npx skills add Rimagination/dy-note --skill dy-noteRegistry metadata
This page exposes the same decision, trust, audit, use-case, and install signals through the Registry API, so agents can rank this skill without scraping the UI.
Manifest
/api/registry/manifest/rimagination-dy-note
LLM text
/api/registry/manifest/rimagination-dy-note?format=text
Install alias
/api/registry/install/rimagination-dy-note
Recommend
/api/registry/recommend?task=Use%20dy-note%20in%20an%20agent%20workflow&limit=3
Agent fit
Research agents
Use-case tags
Platforms
Claude Code, OpenAI Agents, Browser agents
Audit report
A machine-readable review of install readiness, security metadata, maintenance, and adoption risk.
Agent decision cockpit
Prototype with this skill first; keep a fallback candidate ready.
Role in stack
Fallback candidate
Primary fit
Research agents
Trust label
Prototype first
Install path
Command ready
Use when
Evidence
review first
Implementation path
Trust profile
Trust Score v5 found insufficient evidence for agent installation. Treat this as discovery material, not an executable recommendation.
GitHub adoption
INFO159 GitHub stars
Stars/forks activity
CHECK159 stars, 24 forks; issue activity unavailable in current metadata
Recent maintenance
PASS2mo since push
License clarity
PASSMIT
Good signals
Review before install
Recommended action
Choose a stronger alternative or inspect the source manually before any install attempt.
Quality profile
Useful candidate, but compare it with alternatives before adopting.
Workflow fit
Investigate faster
I need my agent to research a topic, compare sources, and produce a concise report.
Publish consistently
I need my agent to turn research and product updates into useful content drafts.
Process rich media
I need my agent to process images, video, or audio and extract useful information.
Workflow fit
Find, compare, and synthesize
A workflow for agents that gather sources, compare claims, summarize long material, and draft useful research briefs.
Turn skills into distribution
A workflow for turning newly indexed skills into SEO briefs, social drafts, comparison pages, and reusable publishing workflows.
Scrape, clean, and reuse web data
A practical workflow for agents that crawl public pages, extract clean content, normalize data, and hand it to downstream research or RAG workflows.
Alternative shortlist
Similar skills that may fit this task.
Generate original one-ink or controlled two-ink editorial images from any theme, sentence, article idea, object, or reference photo. Always use this skill when the user asks for 单色海报、双色印刷、单色调视觉、蓝色/绿色孔版印刷、risograph、网点照片、复古或当代编辑排版、zine poster, monochrome editorial poster, duotone print, or asks to use the mono-color style. It uses an adaptive white, gray, or pale-beige substrate, no more than two printing inks, active negative space, terse human language, and strong serif/grotesk/mono typography without making retro styling the default or copying a source composition, wording, logo, or artwork. Produce both the final generation prompt and the generated raster image unless the user explicitly asks for prompt only.
Research the last 30 days across Reddit, X, YouTube, Hacker News, Polymarket, GitHub, and the web, then synthesize a grounded brief for an AI agent.
Academic Research Skills for Claude Code: research → write → review → revise → finalize
A relentless interview to sharpen a plan or design.
--- name: dy-note description: "DyNote: systematically and efficiently extract raw Douyin/DY video data and analyze videos, comments, accounts, hashtags, and short-video scenes into evidence-graded learning notes, summaries, research briefs, scripts, and knowledge-base material. Use when the user asks to 抓取/提取/整理 抖音视频字幕、视频文案、ASR 转写、Qwen3-ASR 中文转写、原始材料归档、学习笔记、analysis plan、note budget、避免返工、复用已有素材、评论洞察、账号分析、赛道/话题研究、竞品拆解、电商/本地生活视频分析、事实核查、自动搜索素材, save Douyin content as Markdown/TXT, or use subtitle/local ASR as the factual spine with logged-in Douyin Web built-in AI / Doubao fallback as visual or quick-reading supplements." ---
# DyNote
DyNote 是面向 Codex 这类 Agent 的抖音学习工具,不是一次性摘要器。核心原则是“数据资产先行,学习笔记后置”:先把字幕/转写、评论、元数据和 AI 快读沉淀为可复用资产,再按用户需求生成可追溯的学习笔记、总结和写作材料。默认目标不是字幕工程文件,而是先落一份原始数据包:`douyin_ai_brief.md`、`douyin_ai_brief.json`、`transcript.cleaned.md`、`transcript.txt`、`segments.json`、`metadata.json`、`note_budget.json`,并用 `assets/` 归档可复用资产。默认把独立字幕轨或本地自动语音识别转写当作事实主干;当转写密度低、任务需要画面理解或用户只要快速筛选时,再用已登录抖音网页版的“问AI / 识别画面”补充,豆包只作为抖音 AI 不可用时的备用快读或待核验假设。
联网或登录态操作必须先使用 `web-access`。不要读取、复制或打印 Cookie、msToken、a_bogus、x-secsdk-web-signature、临时签名视频 URL 等敏感参数;脚本只让已授权 Chrome 页面自己加载内容。
DyNote 与 Bili Note 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。
## 浏览器与登录态硬规则
- 抖音内置 AI 和豆包备用路线都只使用 `web-access` 连接到用户当前可用的 Chrome。不要启动无登录态 Playwright 浏览器,不要用静态 curl 抓登录页,不要导出或保存 `storageState`、Cookie、localStorage 或 token。 - 抖音内置 AI 路线需要当前 Chrome 已登录抖音网页版。它主要用于低转写密度、画面文字、镜头/场景或快速筛选;打开视频后使用页面右侧 `问AI`,必要时点击可见的 `识别画面` 把当前帧加入问答上下文。 - 如果抖音页面没有 `问AI` / `识别画面` 或未生成 `章节要点`,记录为 `weak` 或 `blocked`,先确保字幕/本地自动语音识别事实主干可用,再考虑豆包备用、关键帧或 OCR。 - 在向豆包发送内容前,必须确认 `https://www.doubao.com/chat/` 在当前 Chrome 中已登录且有可见聊天输入框、侧边栏/新对话等用户态界面。 - 如果未检测到豆包登录态,停止并返回 `blocked: doubao-login-required`,提示用户先在同一个 Chrome 登录豆包。不要静默降级到其他浏览器。 - 豆包备用快速解读优先使用用户复制的完整抖音分享文本,不要只喂最终 `douyin.com/video/...`,因为完整分享文案更容易触发豆包的搜索/参考资料式视频概述。 - 豆包输出需要做证据分级:`search-derived` 是检索式概述,`visual-claimed` 是声称包含画面/镜头细节,`blocked` 是豆包无法访问视频画面,`weak` 是信息不足。不要把检索式概述说成逐帧视觉解析,也不要让豆包替代完整字幕或本地转写。
## 抖音字幕现实与默认路线
抖音和 B 站不同:很多抖音视频没有可直接抓取的独立字幕文件。先按用户任务分流,不要固定把所有视频都下载转写。
- 用户只是问“这个视频讲什么”、想做选题筛选、草稿或快速理解时,可以走已登录抖音网页版内置 AI 快读;但输出必须标注为快读/视觉假设,不写成完整字幕提取。已有 `douyin_ai_brief.json` 时先复用;如果不可用,再用豆包 `fast` 备用。 - 用户要学习笔记、可靠原文、逐句内容、引用、脚本拆解、事实核查或可发布材料时,先找已有 SRT/VTT/TXT;没有可用字幕轨或转写时,主要依赖本地自动语音识别。中文或未指定语言优先共享 Qwen3-ASR,明确外语视频再用 Whisper 系后端。 - 用户要镜头、画面文字、贴纸文字、操作步骤、商品/价格/场景细节,或 `note_budget.json` 显示转写密度低时,不能只靠音频转写;优先补抖音 `问AI / 识别画面`、关键帧、截图或 OCR。抖音 AI 仍不能提取完整字幕;豆包 `evidence` 只作为抖音 AI 不行时的备用视觉假设。
抖音字幕常见两种形态:
- 独立字幕轨:创作者使用平台字幕功能或上传 SRT 后,网页播放器可能叠加渲染 VTT 字幕。若能抓到这类轨道,可作为逐句文本材料。 - 画面内嵌文字:字幕、贴纸或手动排版文字已经焊在画面里,没有独立文件。音频转写只能识别人声,不能读取这类画面文字,必须补视觉证据。
如果视频较长,但 `note_budget.json` 中 `visual_dependency.risk` 为 `medium` 或 `high`,必须提醒用户:转写文本过少,完整理解可能依赖画面,不能把稀疏本地自动语音识别结果写成完整笔记。抖音问 AI、豆包和关键帧结果必须作为补充证据审计,不能替代原文主干。
## 场景模式路由
先判断用户真正要完成的任务,再决定证据深度和工具路线:
- `single-video-note`:默认模式。单条视频/分享文本 -> 字幕/本地自动语音识别做事实主干;转写稀疏或任务需要画面时补抖音问 AI、关键帧/OCR;抖音 AI 不可用时豆包才作为备用假设,输出可读笔记。 - `comment-insight`:用户关心评论、痛点、需求、FAQ、反对意见或爆点反馈时,加载 `douyin-comments`。默认只抓前 100 条主评论及这些主评论的楼中楼,输出 `_sample.json/csv`,并明确提示这不是全部评论。用户明确要完整评论资产、复核全部可见评论或样本不足时,再用 `--full` 做全量抓取。抓到的 JSON/CSV 必须归档到 `assets/comments/`,再输出用户洞察,而不是只把评论写进一次性总结。 - `account-analysis`:账号主页或多条视频 -> 定位、内容支柱、钩子模板、系列化栏目、发布节奏和可复用选题。 - `topic-research`:话题、关键词、赛道、竞品或“自动搜索” -> 先低成本收集标题/简介/话题/样本链接,再按需要升级到 ASR、评论和关键帧。 - `script-mining`:拆脚本、镜头、叙事节奏、开头钩子、转场、结尾 CTA;脚本文案以字幕/本地转写为主,抖音内置 AI 或备用豆包只给画面假设,重要结论要用转写/抽帧校验。 - `commerce-analysis`:带货、本地生活、探店、课程或服务视频 -> 卖点、信任证据、价格/优惠、CTA、转化阻力和评论需求。 - `fact-check`:涉及医学、法律、投资、新闻或强事实判断时,区分视频原文、抖音内置 AI/豆包概述和外部来源;高风险结论必须联网核验并标注来源。 - `knowledge-archive`:用户要沉淀资料库、Obsidian、RAG 或写作素材时,保留来源 URL、作者、时间、证据等级、关键词和后续可检索标签。
成本分层默认从轻到重:
- `quick-pass`:分享文本、页面元数据、抖音内置 AI,必要时备用豆包 `fast`;适合秒级判断、选题筛选和草稿,但必须标注不是完整字幕/全文证据。 - `evidence-pass`:独立字幕轨或 ASR 全文、关键帧/OCR、评论样本;适合要引用、拆解或发布的内容。 - `research-pass`:批量视频、账号/话题搜索、竞品对比和评论聚类;范围大时先给样本计划和 token/时间风险。
## 系统化分析协议
默认按“问题 -> 取证 -> 分析 -> 审计”推进,不要把工具输出直接等同于结论:
1. `research-question`:写清要回答的问题、分析单位和场景模式。复杂任务先生成 `analysis_plan.json`。 2. `sampling-plan`:账号、话题、评论或竞品任务必须说明样本怎么选、样本量是多少、为什么足够或不足。 3. `evidence-ladder`:把证据分为用户输入、页面元数据、独立字幕轨/本地自动语音识别转写、抖音内置 AI、备用豆包快读、评论、关键帧/OCR、外部来源。字幕/转写是事实主干;抖音 AI 和豆包是快读或视觉补充。结论必须标注依赖哪一层。 4. `synthesis-gate`:合成前先读取 `assets/asset_manifest.json` 或确认同等原始材料,检查证据等级、覆盖范围和反例/不确定性;缺证据时先写范围限制,不要补故事。 5. `audit-trail`:最终笔记或研究简报保留来源 URL、采集时间、输出文件、样本范围、`note_budget.json` 和无法验证的点。
## 高效执行与复用策略
- 先检查已有产物,再决定下一步。已有 `douyin_ai_brief.json` 时,不要重复问抖音内置 AI;已有 `doubao_brief.json` 时,不要重复问豆包;已有 `transcript.txt`、`segments.json`、`metadata.json` 时,不要重跑 ASR;已有 `note_budget.json` 且未过期时,不要重算预算。 - `analysis_plan.json` 只在复杂任务或目标变化时创建;已有计划默认复用。目标、来源、模式或证据等级变化时才用 `--force` 重建。 - 先走最便宜的 `quick-pass`,只有当研究问题无法回答、证据等级不足、或用户要可发布笔记/事实核查时,才升级到 `evidence-pass` 或 `research-pass`。评论区任务的 `quick-pass` 是前 100 条主评论及对应楼中楼样本;全量可见评论属于更重的资产补齐步骤。 - 不要为了“完整流程”固定执行所有步骤。单条视频如果已有高质量转写,可直接预算和写笔记;评论洞察如果只问观众反馈,可以先抓 100 条评论样本,不必先全量 ASR,也不必默认抓完整评论区。 - 重新运行昂贵步骤前必须说明触发条件:输入变了、旧文件缺失/损坏/过期、证据等级不足,或用户明确要求更高质量。
## 原始材料与学习笔记默认策略
- 默认先建立数据资产,再写学习笔记。不要只把抖音内置 AI、豆包概述或未经审计的本地自动语音识别文本直接当最终笔记。 - `douyin_ai_brief.json`、`doubao_brief.json`、`transcript.txt`、`segments.json`、`metadata.json`、评论 JSON/CSV 和关键帧截图都属于原始数据;最终学习笔记必须能回到这些材料解释来源。 - 默认把字幕/转写和完整评论整理成资产包:`assets/transcripts/` 保存字幕、转写和片段;`assets/comments/` 保存完整评论 JSON/CSV、JSONL 明细和可读 Markdown;`assets/asset_manifest.json` 是后续再分析的入口。 - `assets/asset_manifest.json` 是事实入口;`learning_note.md` 是从资产生成的一种学习视图,不是资产本身。用户换问题、换场景或要求复核时,优先复用资产重新组织笔记,不要重跑或覆盖原始材料。 - 写笔记前先确认用户需求:内容学习、脚本复盘、评论洞察、事实核查、写作素材或知识库归档。再从资产中选择证据和结构,不要先脑补结论再找材料。 - 每次生成转写材料后读取 `note_budget.json`。它根据视频时长、转写字数、片段/证据块、评论数和互动质量给出推荐笔记长度。 - 长视频、长转写、高评论或高互动视频要写更长、更结构化的笔记;短视频或低信息密度视频避免过度扩写。 - 互动质量是“值得多写”的辅助信号,不替代证据。扩写必须来自视频原文、评论样本、关键帧、外部核验,或已明确降级为视觉假设的抖音内置 AI/豆包快读。 - 学习型笔记要让读者像学完一个短课题:获得概念、判断标准、方法步骤、适用边界、坑点和可迁移用法。不要只写“视频讲了什么”。
## 默认流程
1. 读清用户要的是哪种场景模式:单条视频笔记、评论洞察、账号分析、话题研究、脚本拆解、电商分析、事实核查,还是知识库归档。 2. 首次使用、换机器、或准备跑完整抖音链接流程时,先检查环境:
```powershell $skill = "$env:USERPROFILE\.codex\skills\dy-note" $py = "python" & $py "$skill\scripts\check_environment.py" ```
3. 如果有输出目录,先运行或心中执行 `inspect_workflow_state.py`,复用已有计划、抖音内置 AI brief、豆包 brief、转写、评论、预算和评分。 4. 评论、账号、话题、竞品、事实核查或批量研究任务先用 `create_analysis_plan.py` 生成 `analysis_plan.json`;已有计划且目标没变时不要重建。单条视频任务也要在脑中执行同样的证据闸门。 5. 如果已经有 SRT、Whisper JSON 或 TXT,优先走本地整理路线,避免重复下载和转写;脚本会生成 `note_budget.json`。 6. 如果用户要求可靠全文、学习笔记、逐句内容、引用、脚本拆解或事实核查,优先使用 `extract_douyin_text.py` 取得字幕轨或本地自动语音识别转写。默认 `--asr-backend auto`:中文或未指定语言优先共享 Qwen3-ASR,明确外语视频优先 Whisper。默认复用已有输出;需要重跑时加 `--force`。 7. 如果用户只是快速理解、选题筛选或先拿草稿,且当前 Chrome 已登录抖音网页版,可以跑 `douyin_web_ai_brief.py`;已有可用 `douyin_ai_brief.json` 时先读旧结果。抖音内置 AI 不可用、弱,或 `note_budget.json` 显示低转写密度时,再用 `doubao_video_brief.py` 的 `fast/evidence` 模式备用,但必须标注为假设。 8. 评论、账号、话题、竞品或批量研究任务先做 `quick-pass` 样本,不要直接对大量视频逐条 ASR;样本结论不足时再升级到 `evidence-pass` 或 `research-pass`。大评论区默认用 `fetch_douyin_comments.py` 抓前 100 条主评论及对应楼中楼;需要完整可见评论时再显式加 `--full`,不要黑盒等待一轮全量抓取。 9. 每次得到转写、字幕、评论 JSON/CSV、抖音 AI brief 或豆包 brief 后,运行 `archive_dy_note_assets.py` 生成或更新 `assets/`。评论区任务必须保留评论样本或完整评论资产,不能只输出评论摘要。 10. 写学习笔记前,先打开 `assets/asset_manifest.json`,再按用户需求读取 `transcript.cleaned.md`、`segments.json`、完整评论、`douyin_ai_brief.md`、`doubao_brief.md` 或 `analysis_plan.json`,检查证据是否足以回答研究问题。再打开 `note_budget.json`,按推荐长度和 `writing_guidance` 写学习型笔记;如果 `visual_dependency.needs_visual_review=true`,必须在笔记和回复中提醒画面证据不足,或先补抖音 `问AI / 识别画面`、关键帧/OCR。对外部 AI 回答逐项做证据审计,不要把它们的扩写混成视频原文。 11. 如果识别出片名、人名、地名明显错,优先在最终说明里标注可疑词;用户要求校对时再做替换、二次 ASR 或补关键帧。 12. 用户明确要 SRT/VTT/时间轴时,才单独交付字幕文件;无论是否单独交付,都要在 `assets/transcripts/` 里保留可复用文本资产。
## 常用命令
### 0. 检查已有工作状态
已有输出目录时先检查状态,决定下一步,不要盲目重跑:
```powershell & $py "$skill\scripts\inspect_workflow_state.py" ` --out-dir ".\dy_note_output" ` --mode "single-video-note" ```
重点看:
- `reusable_artifacts`:可以直接复用的产物。 - `recommended_next_steps`:真正缺的下一步。 - `avoid_rework`:明确不要重复做的昂贵步骤。 - `stale.note_budget` / `stale.note_score`:预算或评分是否因新材料而过期。
### 1. 创建系统化分析计划
复杂任务先生成计划,再采集数据。单条视频可省略显式文件,但账号/话题/评论/竞品/事实核查任务建议保留:
```powershell & $py "$skill\scripts\create_analysis_plan.py" ` --mode "topic-research" ` --tier "quick-pass" ` --objective "分析这个赛道里什么视频形式值得复用" ` --source "铁板牛排 炸土豆饼 野外烹饪" ` --out-dir ".\dy_note_research" ```
如果 `analysis_plan.json` 已存在,脚本默认复用旧计划;只有目标、来源、模式或证据等级改变时才加 `--force` 覆盖。
常用模式:`single-video-note`、`comment-insight`、`account-analysis`、`topic-research`、`script-mining`、`commerce-analysis`、`fact-check`、`knowledge-archive`。
### 2. 从已有 SRT 生成干净文本
```powershell & $py "$skill\scripts\extract_douyin_text.py" ` --from-srt "D:\微信推送\douyin_subtitles_7647145112421633320\7647145112421633320_16k.srt" ` --metadata-json "D:\微信推送\douyin_subtitles_7647145112421633320\official_detail_summary.json" ` --out-dir "D:\微信推送\dy_note_7647145112421633320" ```
### 3. 从 Whisper JSON 或 TXT 生成干净文本
```powershell & $py "$skill\scripts\extract_douyin_text.py" ` --from-whisper-json ".\audio_16k.json" ` --out-dir ".\dy_note_output"
& $py "$skill\scripts\extract_douyin_text.py" ` --from-txt ".\raw_transcript.txt" ` --source-url "https://www.douyin.com/video/..." ` --out-dir ".\dy_note_output" ```
### 4. 安装或检查共享 Qwen3-ASR 本地环境
Qwen3-ASR 是中文视频优先使用的本地自动语音识别后端。首次使用时安装到共享 venv,复用现有 CUDA Torch;DyNote 和 Bili Note 共用同一套环境:
```powershell & $py "$skill\scripts\setup_qwen_asr_env.py" & $py "$skill\scripts\check_environment.py" ```
看到 `routes.qwen3_asr=OK` 后再使用 Qwen 后端。本机默认 venv 路径是:
```text %USERPROFILE%\.cache\rimagination-notes\qwen3-asr-venv ```
为兼容早期原型,脚本仍会探测旧路径 `%USERPROFILE%\.cache\dy-note\qwen3-asr-venv` 和 `%USERPROFILE%\.cache\douyin-note\qwen3-asr-venv`。
### 5. 从抖音链接完整提取
先按 `web-access` 要求启动并检查 CDP proxy,再运行:
```powershell & $py "$skill\scripts\extract_douyin_text.py" ` "https://v.douyin.com/xxxxxxx/" ` --out-dir "D:\微信推送\dy_note_output" ` --asr-model medium ` --language Chinese ```
如果输出目录已有 `transcript.txt`、`segments.json` 和 `metadata.json`,脚本默认复用并跳过浏览器、下载和 ASR。确实要重跑时加:
```powershell --force ```
中文长视频默认优先用 Qwen3-ASR-0.6B。8GB 显存建议保留默认 60 秒分段,避免整段长音频 OOM:
```powershell & $py "$skill\scripts\extract_douyin_text.py" ` "https://v.douyin.com/xxxxxxx/" ` --out-dir "D:\微信推送\dy_note_output_qwen" ` --asr-backend qwen3-asr ` --qwen-model "Qwen/Qwen3-ASR-0.6B" ` --qwen-chunk-seconds 60 ` --language Chinese ```
脚本会输出:
- `transcript.cleaned.md`:适合阅读和继续写笔记的 Markdown。 - `transcript.txt`:纯文本正文,适合喂给总结、RAG 或写作流程。 - `segments.json`:按原始字幕/ASR 片段保留的结构化文本。 - `metadata.json`:来源、作者、作品 ID、片段数、生成时间和输出清单。 - `note_budget.json`:按时长、转写字数、片段数、评论量和互动质量生成的推荐学习笔记长度,以及转写过稀时的画面依赖提示。 - `page_metadata.json`、视频、音频、Whisper SRT 或 Qwen JSON:完整流程产生的中间材料,供排错和回查使用。
### 6. 从已有音频使用 Qwen 转写
```powershell & $py "$skill\scripts\extract_douyin_text.py" ` --from-audio "D:\微信推送\video_16k.wav" ` --asr-backend qwen3-asr ` --qwen-chunk-seconds 60 ` --metadata-json ".\metadata.json" ` --out-dir ".\dy_note_qwen" ```
### 7. 复用已打开的 web-access target
如果已经用 `web-access` 打开抖音页面并拿到 target id:
```powershell & $py "$skill\scripts\extract_douyin_text.py" ` "https://www.douyin.com/video/7647145112421633320" ` --target "CDP_TARGET_ID" ` --keep-tab ` --out-dir ".\dy_note_output" ```
不要关闭用户已有 tab;只有脚本自己新建的 tab 可以自动关闭。
### 8. 用抖音内置 AI 快速解读视频
先按 `web-access` 要求启动并检查 CDP proxy。脚本会使用当前 Chrome 打开抖音视频页,优先读取页面 `问AI` 生成的 `章节要点` 和时间线;如果传入的是 `jingxuan?modal_id=...`,会自动归一到更稳定的 `/video/<id>` 页面:
```powershell & $py "$skill\scripts\douyin_web_ai_brief.py" ` "https://www.douyin.com/jingxuan?modal_id=7655645985318085322" ` --out-dir ".\dy_note_douyin_ai_7655645985318085322" ```
如果任务依赖当前画面内容,可尝试把暂停帧加入抖音 AI 输入框:
```powershell & $py "$skill\scripts\douyin_web_ai_brief.py" ` "https://www.douyin.com/jingxuan?modal_id=7655645985318085322" ` --identify-frame ` --out-dir ".\dy_note_douyin_ai_frame" ```
脚本会输出:
- `douyin_ai_brief.md`:抖音内置 AI 的章节要点、时间线、识别画面状态和局限。 - `douyin_ai_brief.json`:来源 URL、归一
Source provenance
Decision snapshot
recent repository activity
Audit
Install and adoption review
Agent-proven evidence
Outcome reports after resolve, review, install, and one narrow run.
No agent outcome data yet. The first agent run can report success, setup needs, risk blocks, failure, or not-relevant through /api/agent/outcome.
Install
Free and open source. Review the report before installing into production agents.
Growth loop
Scenario-led draft for dy-note, ready for a manual X post.
dy-note: DyNote: systematically and efficiently extract raw Douyin/DY video data and analyze videos, c... 159 stars https://www.openagentskill.com/skills/rimagination-dy-note?ref=x
Listing + install path for dy-note: https://www.openagentskill.com/skills/rimagination-dy-note?ref=x Install: npx skills add Rimagination/dy-note --skill dy-note
Listing source
This listing was indexed from public sources and is not marked official until a maintainer claim is approved.
Attribution links to the public repository or creator profile. Creators can claim the listing to update ownership signals.
Claim this skillOwner claim
This Registry indexed listing is attributed to Rimagination but is not marked official yet. Claim it to add a verified owner signal and make future launch, install, and audit updates easier to trust.
Creator backlink kit
Show the canonical listing, current trust and audit signals, and real Agent-Proven evidence where developers evaluate the repository.
[](https://www.openagentskill.com/skills/rimagination-dy-note?ref=github&utm_source=github&utm_medium=referral&utm_campaign=creator_badge)
[](https://www.openagentskill.com/skills/rimagination-dy-note?ref=github&utm_source=github&utm_medium=referral&utm_campaign=creator_badge)
[](https://www.openagentskill.com/skills/rimagination-dy-note/audit)
[](https://www.openagentskill.com/skills/rimagination-dy-note?ref=github&utm_source=github&utm_medium=referral&utm_campaign=creator_badge)Rimagination
@rimagination
Share whether this skill looks useful for your agent workflow. Aggregated feedback improves rankings over time.
Do not auto-install
mono-color
Generate original one-ink or controlled two-ink editorial images from any theme, sentence, article idea, object, or reference photo. Always use this skill when the user asks for 单色海报、双色印刷、单色调视觉、蓝色/绿色孔版印刷、risograph、网点照片、复古或当代编辑排版、zine poster, monochrome editorial poster, duotone print, or asks to use the mono-color style. It uses an adaptive white, gray, or pale-beige substrate, no more than two printing inks, active negative space, terse human language, and strong serif/grotesk/mono typography without making retro styling the default or copying a source composition, wording, logo, or artwork. Produce both the final generation prompt and the generated raster image unless the user explicitly asks for prompt only.
1.9K StarsLast30days Skill
Research the last 30 days across Reddit, X, YouTube, Hacker News, Polymarket, GitHub, and the web, then synthesize a grounded brief for an AI agent.
61.0K StarsAcademic Research Skills
Academic Research Skills for Claude Code: research → write → review → revise → finalize
38.4K Starsgrill-me
A relentless interview to sharpen a plan or design.
256.3K StarsDo not auto-install
Install targets
Codex install prompt
Install the "dy-note" agent skill from https://github.com/Rimagination/dy-note/blob/main/SKILL.md. Read its SKILL.md or equivalent instructions first, install only the files needed for this workspace, and summarize any required setup before using it. Skill purpose: DyNote: systematically and efficiently extract raw Douyin/DY video data and analyze videos, comments, accounts, hashtags, and short-video scenes into evidence-graded learning notes, summaries, research briefs, scripts, and knowledge-base material. Use when the user asks to 抓取/提取/整理 抖音视频字幕、视频文案、ASR 转写、Qwen3-ASR 中文转写、原始材料归档、学习笔记、analysis plan、note budget、避免返工、复用已有素材、评论洞察、账号分析、赛道/话题研究、竞品拆解、电商/本地生活视频分析、事实核查、自动搜索素材, save Douyin content as Markdown/TXT, or use subtitle/local ASR as the factual spine with logged-in Douyin Web built-in AI / Doubao fallback as visual or quick-reading supplements. After the install attempt, report the result to https://www.openagentskill.com/api/agent/outcome with POST JSON {"event_id":"install_<unique-id>","skill_slug":"rimagination-dy-note","task":"Install dy-note","agent":"codex","outcome":"success","install_used":true}. Replace event_id with a unique value and outcome with success or failed. Report success only after the skill is installed and a minimal verification passes.Supply asset profile
Deep research, source comparison, literature review, RAG, knowledge search, and reports.
Scenario
Research agents
I need my agent to research a topic, compare sources, and produce a concise report.
Agent fit
Claude Code + OpenAI Agents + Browser agents
Codex, Claude Code, Cursor, CLI, or custom agents.
Install
Ready
npx skills add Rimagination/dy-note --skill dy-note
Maintenance
active
2mo since push
Risk
Needs review
Dependency or permission surface needs review
GitHub quality
159
63/100 Quality · 63/100 Trust
Coverage tags
Review notes
Dependency or permission surface needs review · Permission surface may require sandboxing
Agent adoption scorecard
These scores combine public repository metadata, OpenAgentSkill review signals, maintenance freshness, and install readiness. They are a shortlist signal, not a replacement for human review.
Quality
PromisingUseful candidate, but compare it with alternatives before adopting.
Trust
Do not auto-installTrust Score v5 found insufficient evidence for agent installation. Treat this as discovery material, not an executable recommendation.
Audit
Needs reviewA machine-readable review of install readiness, security metadata, maintenance, and adoption risk.
OpenAgentSkill Trust Score v5
Choose a stronger alternative or inspect the source manually before any install attempt.
Stars
159 GitHub stars
Repo activity
159 stars, 24 forks
Maintenance
2mo since push
License
MIT
Install
npx skills add Rimagination/dy-note --skill dy-note
Install safety
Agent-readable metadata
Use this block or the embedded JSON to decide whether an agent should install this skill, choose an alternative, or ask for human review first.
Suited tasks
Suited agents
Install decision
Trust and risk
Outcome loop
Install command
npx skills add Rimagination/dy-note --skill dy-noteDo not use when
Alternative
1.9K Stars
npx skills add yanliudesign/mono-color-skill --skill mono-color
Alternative
61.0K Stars
npx skills add mvanhorn/last30days-skill -g
Alternative
38.4K Stars
npx skills add Imbad0202/academic-research-skills
Alternative
256.3K Stars
npx skills add mattpocock/skills --skill grill-me
Agent safety v2
This skill should not be selected by an agent without explicit human security review.
Do not auto-install. Inspect the source, dependencies, and permission surface first.
high
Skill metadata references terminal, CLI, shell, subprocess, or command execution workflows.
medium
Skill may drive a browser or interact with web pages.
medium
Skill likely fetches remote pages, APIs, repositories, or external services.
medium
Skill may read or write project files, documents, generated artifacts, or local workspace state.
Agent resolve plan
The Resolve API returns the selected skill, alternatives, safety policy, audit notes, install target, and copy-paste prompt an agent can follow without scraping this page.
Open JSON
/api/agent/resolve?task=Use%20dy-note%20for%20an%20agent%20workflow&agent=codex&max_risk=medium
Resolve text
/api/agent/resolve?task=Use%20dy-note%20for%20an%20agent%20workflow&agent=codex&max_risk=medium&format=text
Install handoff
/api/skills/rimagination-dy-note/install
Agent should check
Copy prompt
Task: Use dy-note in this workspace.
Resolve first: https://www.openagentskill.com/api/agent/resolve?task=Use%20dy-note%20for%20an%20agent%20workflow&agent=codex&max_risk=medium
Review install handoff: https://www.openagentskill.com/api/skills/rimagination-dy-note/install
Install command: npx skills add Rimagination/dy-note --skill dy-note
Before running it, summarize audit warnings, required permissions, and the fallback skill if install is risky.Agent handoff
Use the public install endpoint to fetch the command, safety checklist, target prompts, and canonical links for this skill.
Install handoff
/api/skills/rimagination-dy-note/install
LLM text format
/api/skills/rimagination-dy-note/install?format=text
Find alternatives
/api/skills/search?q=dy-note&limit=3
Agent prompt
Use dy-note for this task. Review https://www.openagentskill.com/api/skills/rimagination-dy-note/install, then install with: npx skills add Rimagination/dy-note --skill dy-noteRegistry metadata
This page exposes the same decision, trust, audit, use-case, and install signals through the Registry API, so agents can rank this skill without scraping the UI.
Manifest
/api/registry/manifest/rimagination-dy-note
LLM text
/api/registry/manifest/rimagination-dy-note?format=text
Install alias
/api/registry/install/rimagination-dy-note
Recommend
/api/registry/recommend?task=Use%20dy-note%20in%20an%20agent%20workflow&limit=3
Agent fit
Research agents
Use-case tags
Platforms
Claude Code, OpenAI Agents, Browser agents
Audit report
A machine-readable review of install readiness, security metadata, maintenance, and adoption risk.
Agent decision cockpit
Prototype with this skill first; keep a fallback candidate ready.
Role in stack
Fallback candidate
Primary fit
Research agents
Trust label
Prototype first
Install path
Command ready
Use when
Evidence
review first
Implementation path
Trust profile
Trust Score v5 found insufficient evidence for agent installation. Treat this as discovery material, not an executable recommendation.
GitHub adoption
INFO159 GitHub stars
Stars/forks activity
CHECK159 stars, 24 forks; issue activity unavailable in current metadata
Recent maintenance
PASS2mo since push
License clarity
PASSMIT
Good signals
Review before install
Recommended action
Choose a stronger alternative or inspect the source manually before any install attempt.
Quality profile
Useful candidate, but compare it with alternatives before adopting.
Workflow fit
Investigate faster
I need my agent to research a topic, compare sources, and produce a concise report.
Publish consistently
I need my agent to turn research and product updates into useful content drafts.
Process rich media
I need my agent to process images, video, or audio and extract useful information.
Workflow fit
Find, compare, and synthesize
A workflow for agents that gather sources, compare claims, summarize long material, and draft useful research briefs.
Turn skills into distribution
A workflow for turning newly indexed skills into SEO briefs, social drafts, comparison pages, and reusable publishing workflows.
Scrape, clean, and reuse web data
A practical workflow for agents that crawl public pages, extract clean content, normalize data, and hand it to downstream research or RAG workflows.
Alternative shortlist
Similar skills that may fit this task.
Generate original one-ink or controlled two-ink editorial images from any theme, sentence, article idea, object, or reference photo. Always use this skill when the user asks for 单色海报、双色印刷、单色调视觉、蓝色/绿色孔版印刷、risograph、网点照片、复古或当代编辑排版、zine poster, monochrome editorial poster, duotone print, or asks to use the mono-color style. It uses an adaptive white, gray, or pale-beige substrate, no more than two printing inks, active negative space, terse human language, and strong serif/grotesk/mono typography without making retro styling the default or copying a source composition, wording, logo, or artwork. Produce both the final generation prompt and the generated raster image unless the user explicitly asks for prompt only.
Research the last 30 days across Reddit, X, YouTube, Hacker News, Polymarket, GitHub, and the web, then synthesize a grounded brief for an AI agent.
Academic Research Skills for Claude Code: research → write → review → revise → finalize
A relentless interview to sharpen a plan or design.
--- name: dy-note description: "DyNote: systematically and efficiently extract raw Douyin/DY video data and analyze videos, comments, accounts, hashtags, and short-video scenes into evidence-graded learning notes, summaries, research briefs, scripts, and knowledge-base material. Use when the user asks to 抓取/提取/整理 抖音视频字幕、视频文案、ASR 转写、Qwen3-ASR 中文转写、原始材料归档、学习笔记、analysis plan、note budget、避免返工、复用已有素材、评论洞察、账号分析、赛道/话题研究、竞品拆解、电商/本地生活视频分析、事实核查、自动搜索素材, save Douyin content as Markdown/TXT, or use subtitle/local ASR as the factual spine with logged-in Douyin Web built-in AI / Doubao fallback as visual or quick-reading supplements." ---
# DyNote
DyNote 是面向 Codex 这类 Agent 的抖音学习工具,不是一次性摘要器。核心原则是“数据资产先行,学习笔记后置”:先把字幕/转写、评论、元数据和 AI 快读沉淀为可复用资产,再按用户需求生成可追溯的学习笔记、总结和写作材料。默认目标不是字幕工程文件,而是先落一份原始数据包:`douyin_ai_brief.md`、`douyin_ai_brief.json`、`transcript.cleaned.md`、`transcript.txt`、`segments.json`、`metadata.json`、`note_budget.json`,并用 `assets/` 归档可复用资产。默认把独立字幕轨或本地自动语音识别转写当作事实主干;当转写密度低、任务需要画面理解或用户只要快速筛选时,再用已登录抖音网页版的“问AI / 识别画面”补充,豆包只作为抖音 AI 不可用时的备用快读或待核验假设。
联网或登录态操作必须先使用 `web-access`。不要读取、复制或打印 Cookie、msToken、a_bogus、x-secsdk-web-signature、临时签名视频 URL 等敏感参数;脚本只让已授权 Chrome 页面自己加载内容。
DyNote 与 Bili Note 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。
## 浏览器与登录态硬规则
- 抖音内置 AI 和豆包备用路线都只使用 `web-access` 连接到用户当前可用的 Chrome。不要启动无登录态 Playwright 浏览器,不要用静态 curl 抓登录页,不要导出或保存 `storageState`、Cookie、localStorage 或 token。 - 抖音内置 AI 路线需要当前 Chrome 已登录抖音网页版。它主要用于低转写密度、画面文字、镜头/场景或快速筛选;打开视频后使用页面右侧 `问AI`,必要时点击可见的 `识别画面` 把当前帧加入问答上下文。 - 如果抖音页面没有 `问AI` / `识别画面` 或未生成 `章节要点`,记录为 `weak` 或 `blocked`,先确保字幕/本地自动语音识别事实主干可用,再考虑豆包备用、关键帧或 OCR。 - 在向豆包发送内容前,必须确认 `https://www.doubao.com/chat/` 在当前 Chrome 中已登录且有可见聊天输入框、侧边栏/新对话等用户态界面。 - 如果未检测到豆包登录态,停止并返回 `blocked: doubao-login-required`,提示用户先在同一个 Chrome 登录豆包。不要静默降级到其他浏览器。 - 豆包备用快速解读优先使用用户复制的完整抖音分享文本,不要只喂最终 `douyin.com/video/...`,因为完整分享文案更容易触发豆包的搜索/参考资料式视频概述。 - 豆包输出需要做证据分级:`search-derived` 是检索式概述,`visual-claimed` 是声称包含画面/镜头细节,`blocked` 是豆包无法访问视频画面,`weak` 是信息不足。不要把检索式概述说成逐帧视觉解析,也不要让豆包替代完整字幕或本地转写。
## 抖音字幕现实与默认路线
抖音和 B 站不同:很多抖音视频没有可直接抓取的独立字幕文件。先按用户任务分流,不要固定把所有视频都下载转写。
- 用户只是问“这个视频讲什么”、想做选题筛选、草稿或快速理解时,可以走已登录抖音网页版内置 AI 快读;但输出必须标注为快读/视觉假设,不写成完整字幕提取。已有 `douyin_ai_brief.json` 时先复用;如果不可用,再用豆包 `fast` 备用。 - 用户要学习笔记、可靠原文、逐句内容、引用、脚本拆解、事实核查或可发布材料时,先找已有 SRT/VTT/TXT;没有可用字幕轨或转写时,主要依赖本地自动语音识别。中文或未指定语言优先共享 Qwen3-ASR,明确外语视频再用 Whisper 系后端。 - 用户要镜头、画面文字、贴纸文字、操作步骤、商品/价格/场景细节,或 `note_budget.json` 显示转写密度低时,不能只靠音频转写;优先补抖音 `问AI / 识别画面`、关键帧、截图或 OCR。抖音 AI 仍不能提取完整字幕;豆包 `evidence` 只作为抖音 AI 不行时的备用视觉假设。
抖音字幕常见两种形态:
- 独立字幕轨:创作者使用平台字幕功能或上传 SRT 后,网页播放器可能叠加渲染 VTT 字幕。若能抓到这类轨道,可作为逐句文本材料。 - 画面内嵌文字:字幕、贴纸或手动排版文字已经焊在画面里,没有独立文件。音频转写只能识别人声,不能读取这类画面文字,必须补视觉证据。
如果视频较长,但 `note_budget.json` 中 `visual_dependency.risk` 为 `medium` 或 `high`,必须提醒用户:转写文本过少,完整理解可能依赖画面,不能把稀疏本地自动语音识别结果写成完整笔记。抖音问 AI、豆包和关键帧结果必须作为补充证据审计,不能替代原文主干。
## 场景模式路由
先判断用户真正要完成的任务,再决定证据深度和工具路线:
- `single-video-note`:默认模式。单条视频/分享文本 -> 字幕/本地自动语音识别做事实主干;转写稀疏或任务需要画面时补抖音问 AI、关键帧/OCR;抖音 AI 不可用时豆包才作为备用假设,输出可读笔记。 - `comment-insight`:用户关心评论、痛点、需求、FAQ、反对意见或爆点反馈时,加载 `douyin-comments`。默认只抓前 100 条主评论及这些主评论的楼中楼,输出 `_sample.json/csv`,并明确提示这不是全部评论。用户明确要完整评论资产、复核全部可见评论或样本不足时,再用 `--full` 做全量抓取。抓到的 JSON/CSV 必须归档到 `assets/comments/`,再输出用户洞察,而不是只把评论写进一次性总结。 - `account-analysis`:账号主页或多条视频 -> 定位、内容支柱、钩子模板、系列化栏目、发布节奏和可复用选题。 - `topic-research`:话题、关键词、赛道、竞品或“自动搜索” -> 先低成本收集标题/简介/话题/样本链接,再按需要升级到 ASR、评论和关键帧。 - `script-mining`:拆脚本、镜头、叙事节奏、开头钩子、转场、结尾 CTA;脚本文案以字幕/本地转写为主,抖音内置 AI 或备用豆包只给画面假设,重要结论要用转写/抽帧校验。 - `commerce-analysis`:带货、本地生活、探店、课程或服务视频 -> 卖点、信任证据、价格/优惠、CTA、转化阻力和评论需求。 - `fact-check`:涉及医学、法律、投资、新闻或强事实判断时,区分视频原文、抖音内置 AI/豆包概述和外部来源;高风险结论必须联网核验并标注来源。 - `knowledge-archive`:用户要沉淀资料库、Obsidian、RAG 或写作素材时,保留来源 URL、作者、时间、证据等级、关键词和后续可检索标签。
成本分层默认从轻到重:
- `quick-pass`:分享文本、页面元数据、抖音内置 AI,必要时备用豆包 `fast`;适合秒级判断、选题筛选和草稿,但必须标注不是完整字幕/全文证据。 - `evidence-pass`:独立字幕轨或 ASR 全文、关键帧/OCR、评论样本;适合要引用、拆解或发布的内容。 - `research-pass`:批量视频、账号/话题搜索、竞品对比和评论聚类;范围大时先给样本计划和 token/时间风险。
## 系统化分析协议
默认按“问题 -> 取证 -> 分析 -> 审计”推进,不要把工具输出直接等同于结论:
1. `research-question`:写清要回答的问题、分析单位和场景模式。复杂任务先生成 `analysis_plan.json`。 2. `sampling-plan`:账号、话题、评论或竞品任务必须说明样本怎么选、样本量是多少、为什么足够或不足。 3. `evidence-ladder`:把证据分为用户输入、页面元数据、独立字幕轨/本地自动语音识别转写、抖音内置 AI、备用豆包快读、评论、关键帧/OCR、外部来源。字幕/转写是事实主干;抖音 AI 和豆包是快读或视觉补充。结论必须标注依赖哪一层。 4. `synthesis-gate`:合成前先读取 `assets/asset_manifest.json` 或确认同等原始材料,检查证据等级、覆盖范围和反例/不确定性;缺证据时先写范围限制,不要补故事。 5. `audit-trail`:最终笔记或研究简报保留来源 URL、采集时间、输出文件、样本范围、`note_budget.json` 和无法验证的点。
## 高效执行与复用策略
- 先检查已有产物,再决定下一步。已有 `douyin_ai_brief.json` 时,不要重复问抖音内置 AI;已有 `doubao_brief.json` 时,不要重复问豆包;已有 `transcript.txt`、`segments.json`、`metadata.json` 时,不要重跑 ASR;已有 `note_budget.json` 且未过期时,不要重算预算。 - `analysis_plan.json` 只在复杂任务或目标变化时创建;已有计划默认复用。目标、来源、模式或证据等级变化时才用 `--force` 重建。 - 先走最便宜的 `quick-pass`,只有当研究问题无法回答、证据等级不足、或用户要可发布笔记/事实核查时,才升级到 `evidence-pass` 或 `research-pass`。评论区任务的 `quick-pass` 是前 100 条主评论及对应楼中楼样本;全量可见评论属于更重的资产补齐步骤。 - 不要为了“完整流程”固定执行所有步骤。单条视频如果已有高质量转写,可直接预算和写笔记;评论洞察如果只问观众反馈,可以先抓 100 条评论样本,不必先全量 ASR,也不必默认抓完整评论区。 - 重新运行昂贵步骤前必须说明触发条件:输入变了、旧文件缺失/损坏/过期、证据等级不足,或用户明确要求更高质量。
## 原始材料与学习笔记默认策略
- 默认先建立数据资产,再写学习笔记。不要只把抖音内置 AI、豆包概述或未经审计的本地自动语音识别文本直接当最终笔记。 - `douyin_ai_brief.json`、`doubao_brief.json`、`transcript.txt`、`segments.json`、`metadata.json`、评论 JSON/CSV 和关键帧截图都属于原始数据;最终学习笔记必须能回到这些材料解释来源。 - 默认把字幕/转写和完整评论整理成资产包:`assets/transcripts/` 保存字幕、转写和片段;`assets/comments/` 保存完整评论 JSON/CSV、JSONL 明细和可读 Markdown;`assets/asset_manifest.json` 是后续再分析的入口。 - `assets/asset_manifest.json` 是事实入口;`learning_note.md` 是从资产生成的一种学习视图,不是资产本身。用户换问题、换场景或要求复核时,优先复用资产重新组织笔记,不要重跑或覆盖原始材料。 - 写笔记前先确认用户需求:内容学习、脚本复盘、评论洞察、事实核查、写作素材或知识库归档。再从资产中选择证据和结构,不要先脑补结论再找材料。 - 每次生成转写材料后读取 `note_budget.json`。它根据视频时长、转写字数、片段/证据块、评论数和互动质量给出推荐笔记长度。 - 长视频、长转写、高评论或高互动视频要写更长、更结构化的笔记;短视频或低信息密度视频避免过度扩写。 - 互动质量是“值得多写”的辅助信号,不替代证据。扩写必须来自视频原文、评论样本、关键帧、外部核验,或已明确降级为视觉假设的抖音内置 AI/豆包快读。 - 学习型笔记要让读者像学完一个短课题:获得概念、判断标准、方法步骤、适用边界、坑点和可迁移用法。不要只写“视频讲了什么”。
## 默认流程
1. 读清用户要的是哪种场景模式:单条视频笔记、评论洞察、账号分析、话题研究、脚本拆解、电商分析、事实核查,还是知识库归档。 2. 首次使用、换机器、或准备跑完整抖音链接流程时,先检查环境:
```powershell $skill = "$env:USERPROFILE\.codex\skills\dy-note" $py = "python" & $py "$skill\scripts\check_environment.py" ```
3. 如果有输出目录,先运行或心中执行 `inspect_workflow_state.py`,复用已有计划、抖音内置 AI brief、豆包 brief、转写、评论、预算和评分。 4. 评论、账号、话题、竞品、事实核查或批量研究任务先用 `create_analysis_plan.py` 生成 `analysis_plan.json`;已有计划且目标没变时不要重建。单条视频任务也要在脑中执行同样的证据闸门。 5. 如果已经有 SRT、Whisper JSON 或 TXT,优先走本地整理路线,避免重复下载和转写;脚本会生成 `note_budget.json`。 6. 如果用户要求可靠全文、学习笔记、逐句内容、引用、脚本拆解或事实核查,优先使用 `extract_douyin_text.py` 取得字幕轨或本地自动语音识别转写。默认 `--asr-backend auto`:中文或未指定语言优先共享 Qwen3-ASR,明确外语视频优先 Whisper。默认复用已有输出;需要重跑时加 `--force`。 7. 如果用户只是快速理解、选题筛选或先拿草稿,且当前 Chrome 已登录抖音网页版,可以跑 `douyin_web_ai_brief.py`;已有可用 `douyin_ai_brief.json` 时先读旧结果。抖音内置 AI 不可用、弱,或 `note_budget.json` 显示低转写密度时,再用 `doubao_video_brief.py` 的 `fast/evidence` 模式备用,但必须标注为假设。 8. 评论、账号、话题、竞品或批量研究任务先做 `quick-pass` 样本,不要直接对大量视频逐条 ASR;样本结论不足时再升级到 `evidence-pass` 或 `research-pass`。大评论区默认用 `fetch_douyin_comments.py` 抓前 100 条主评论及对应楼中楼;需要完整可见评论时再显式加 `--full`,不要黑盒等待一轮全量抓取。 9. 每次得到转写、字幕、评论 JSON/CSV、抖音 AI brief 或豆包 brief 后,运行 `archive_dy_note_assets.py` 生成或更新 `assets/`。评论区任务必须保留评论样本或完整评论资产,不能只输出评论摘要。 10. 写学习笔记前,先打开 `assets/asset_manifest.json`,再按用户需求读取 `transcript.cleaned.md`、`segments.json`、完整评论、`douyin_ai_brief.md`、`doubao_brief.md` 或 `analysis_plan.json`,检查证据是否足以回答研究问题。再打开 `note_budget.json`,按推荐长度和 `writing_guidance` 写学习型笔记;如果 `visual_dependency.needs_visual_review=true`,必须在笔记和回复中提醒画面证据不足,或先补抖音 `问AI / 识别画面`、关键帧/OCR。对外部 AI 回答逐项做证据审计,不要把它们的扩写混成视频原文。 11. 如果识别出片名、人名、地名明显错,优先在最终说明里标注可疑词;用户要求校对时再做替换、二次 ASR 或补关键帧。 12. 用户明确要 SRT/VTT/时间轴时,才单独交付字幕文件;无论是否单独交付,都要在 `assets/transcripts/` 里保留可复用文本资产。
## 常用命令
### 0. 检查已有工作状态
已有输出目录时先检查状态,决定下一步,不要盲目重跑:
```powershell & $py "$skill\scripts\inspect_workflow_state.py" ` --out-dir ".\dy_note_output" ` --mode "single-video-note" ```
重点看:
- `reusable_artifacts`:可以直接复用的产物。 - `recommended_next_steps`:真正缺的下一步。 - `avoid_rework`:明确不要重复做的昂贵步骤。 - `stale.note_budget` / `stale.note_score`:预算或评分是否因新材料而过期。
### 1. 创建系统化分析计划
复杂任务先生成计划,再采集数据。单条视频可省略显式文件,但账号/话题/评论/竞品/事实核查任务建议保留:
```powershell & $py "$skill\scripts\create_analysis_plan.py" ` --mode "topic-research" ` --tier "quick-pass" ` --objective "分析这个赛道里什么视频形式值得复用" ` --source "铁板牛排 炸土豆饼 野外烹饪" ` --out-dir ".\dy_note_research" ```
如果 `analysis_plan.json` 已存在,脚本默认复用旧计划;只有目标、来源、模式或证据等级改变时才加 `--force` 覆盖。
常用模式:`single-video-note`、`comment-insight`、`account-analysis`、`topic-research`、`script-mining`、`commerce-analysis`、`fact-check`、`knowledge-archive`。
### 2. 从已有 SRT 生成干净文本
```powershell & $py "$skill\scripts\extract_douyin_text.py" ` --from-srt "D:\微信推送\douyin_subtitles_7647145112421633320\7647145112421633320_16k.srt" ` --metadata-json "D:\微信推送\douyin_subtitles_7647145112421633320\official_detail_summary.json" ` --out-dir "D:\微信推送\dy_note_7647145112421633320" ```
### 3. 从 Whisper JSON 或 TXT 生成干净文本
```powershell & $py "$skill\scripts\extract_douyin_text.py" ` --from-whisper-json ".\audio_16k.json" ` --out-dir ".\dy_note_output"
& $py "$skill\scripts\extract_douyin_text.py" ` --from-txt ".\raw_transcript.txt" ` --source-url "https://www.douyin.com/video/..." ` --out-dir ".\dy_note_output" ```
### 4. 安装或检查共享 Qwen3-ASR 本地环境
Qwen3-ASR 是中文视频优先使用的本地自动语音识别后端。首次使用时安装到共享 venv,复用现有 CUDA Torch;DyNote 和 Bili Note 共用同一套环境:
```powershell & $py "$skill\scripts\setup_qwen_asr_env.py" & $py "$skill\scripts\check_environment.py" ```
看到 `routes.qwen3_asr=OK` 后再使用 Qwen 后端。本机默认 venv 路径是:
```text %USERPROFILE%\.cache\rimagination-notes\qwen3-asr-venv ```
为兼容早期原型,脚本仍会探测旧路径 `%USERPROFILE%\.cache\dy-note\qwen3-asr-venv` 和 `%USERPROFILE%\.cache\douyin-note\qwen3-asr-venv`。
### 5. 从抖音链接完整提取
先按 `web-access` 要求启动并检查 CDP proxy,再运行:
```powershell & $py "$skill\scripts\extract_douyin_text.py" ` "https://v.douyin.com/xxxxxxx/" ` --out-dir "D:\微信推送\dy_note_output" ` --asr-model medium ` --language Chinese ```
如果输出目录已有 `transcript.txt`、`segments.json` 和 `metadata.json`,脚本默认复用并跳过浏览器、下载和 ASR。确实要重跑时加:
```powershell --force ```
中文长视频默认优先用 Qwen3-ASR-0.6B。8GB 显存建议保留默认 60 秒分段,避免整段长音频 OOM:
```powershell & $py "$skill\scripts\extract_douyin_text.py" ` "https://v.douyin.com/xxxxxxx/" ` --out-dir "D:\微信推送\dy_note_output_qwen" ` --asr-backend qwen3-asr ` --qwen-model "Qwen/Qwen3-ASR-0.6B" ` --qwen-chunk-seconds 60 ` --language Chinese ```
脚本会输出:
- `transcript.cleaned.md`:适合阅读和继续写笔记的 Markdown。 - `transcript.txt`:纯文本正文,适合喂给总结、RAG 或写作流程。 - `segments.json`:按原始字幕/ASR 片段保留的结构化文本。 - `metadata.json`:来源、作者、作品 ID、片段数、生成时间和输出清单。 - `note_budget.json`:按时长、转写字数、片段数、评论量和互动质量生成的推荐学习笔记长度,以及转写过稀时的画面依赖提示。 - `page_metadata.json`、视频、音频、Whisper SRT 或 Qwen JSON:完整流程产生的中间材料,供排错和回查使用。
### 6. 从已有音频使用 Qwen 转写
```powershell & $py "$skill\scripts\extract_douyin_text.py" ` --from-audio "D:\微信推送\video_16k.wav" ` --asr-backend qwen3-asr ` --qwen-chunk-seconds 60 ` --metadata-json ".\metadata.json" ` --out-dir ".\dy_note_qwen" ```
### 7. 复用已打开的 web-access target
如果已经用 `web-access` 打开抖音页面并拿到 target id:
```powershell & $py "$skill\scripts\extract_douyin_text.py" ` "https://www.douyin.com/video/7647145112421633320" ` --target "CDP_TARGET_ID" ` --keep-tab ` --out-dir ".\dy_note_output" ```
不要关闭用户已有 tab;只有脚本自己新建的 tab 可以自动关闭。
### 8. 用抖音内置 AI 快速解读视频
先按 `web-access` 要求启动并检查 CDP proxy。脚本会使用当前 Chrome 打开抖音视频页,优先读取页面 `问AI` 生成的 `章节要点` 和时间线;如果传入的是 `jingxuan?modal_id=...`,会自动归一到更稳定的 `/video/<id>` 页面:
```powershell & $py "$skill\scripts\douyin_web_ai_brief.py" ` "https://www.douyin.com/jingxuan?modal_id=7655645985318085322" ` --out-dir ".\dy_note_douyin_ai_7655645985318085322" ```
如果任务依赖当前画面内容,可尝试把暂停帧加入抖音 AI 输入框:
```powershell & $py "$skill\scripts\douyin_web_ai_brief.py" ` "https://www.douyin.com/jingxuan?modal_id=7655645985318085322" ` --identify-frame ` --out-dir ".\dy_note_douyin_ai_frame" ```
脚本会输出:
- `douyin_ai_brief.md`:抖音内置 AI 的章节要点、时间线、识别画面状态和局限。 - `douyin_ai_brief.json`:来源 URL、归一
Source provenance
Decision snapshot
recent repository activity
Audit
Install and adoption review
Agent-proven evidence
Outcome reports after resolve, review, install, and one narrow run.
No agent outcome data yet. The first agent run can report success, setup needs, risk blocks, failure, or not-relevant through /api/agent/outcome.
Install
Free and open source. Review the report before installing into production agents.
Growth loop
Scenario-led draft for dy-note, ready for a manual X post.
dy-note: DyNote: systematically and efficiently extract raw Douyin/DY video data and analyze videos, c... 159 stars https://www.openagentskill.com/skills/rimagination-dy-note?ref=x
Listing + install path for dy-note: https://www.openagentskill.com/skills/rimagination-dy-note?ref=x Install: npx skills add Rimagination/dy-note --skill dy-note
Listing source
This listing was indexed from public sources and is not marked official until a maintainer claim is approved.
Attribution links to the public repository or creator profile. Creators can claim the listing to update ownership signals.
Claim this skillOwner claim
This Registry indexed listing is attributed to Rimagination but is not marked official yet. Claim it to add a verified owner signal and make future launch, install, and audit updates easier to trust.
Creator backlink kit
Show the canonical listing, current trust and audit signals, and real Agent-Proven evidence where developers evaluate the repository.
[](https://www.openagentskill.com/skills/rimagination-dy-note?ref=github&utm_source=github&utm_medium=referral&utm_campaign=creator_badge)
[](https://www.openagentskill.com/skills/rimagination-dy-note?ref=github&utm_source=github&utm_medium=referral&utm_campaign=creator_badge)
[](https://www.openagentskill.com/skills/rimagination-dy-note/audit)
[](https://www.openagentskill.com/skills/rimagination-dy-note?ref=github&utm_source=github&utm_medium=referral&utm_campaign=creator_badge)Rimagination
@rimagination
Share whether this skill looks useful for your agent workflow. Aggregated feedback improves rankings over time.
Do not auto-install
mono-color
Generate original one-ink or controlled two-ink editorial images from any theme, sentence, article idea, object, or reference photo. Always use this skill when the user asks for 单色海报、双色印刷、单色调视觉、蓝色/绿色孔版印刷、risograph、网点照片、复古或当代编辑排版、zine poster, monochrome editorial poster, duotone print, or asks to use the mono-color style. It uses an adaptive white, gray, or pale-beige substrate, no more than two printing inks, active negative space, terse human language, and strong serif/grotesk/mono typography without making retro styling the default or copying a source composition, wording, logo, or artwork. Produce both the final generation prompt and the generated raster image unless the user explicitly asks for prompt only.
1.9K StarsLast30days Skill
Research the last 30 days across Reddit, X, YouTube, Hacker News, Polymarket, GitHub, and the web, then synthesize a grounded brief for an AI agent.
61.0K StarsAcademic Research Skills
Academic Research Skills for Claude Code: research → write → review → revise → finalize
38.4K Starsgrill-me
A relentless interview to sharpen a plan or design.
256.3K StarsDo not auto-install
Install targets
Codex install prompt
Install the "dy-note" agent skill from https://github.com/Rimagination/dy-note/blob/main/SKILL.md. Read its SKILL.md or equivalent instructions first, install only the files needed for this workspace, and summarize any required setup before using it. Skill purpose: DyNote: systematically and efficiently extract raw Douyin/DY video data and analyze videos, comments, accounts, hashtags, and short-video scenes into evidence-graded learning notes, summaries, research briefs, scripts, and knowledge-base material. Use when the user asks to 抓取/提取/整理 抖音视频字幕、视频文案、ASR 转写、Qwen3-ASR 中文转写、原始材料归档、学习笔记、analysis plan、note budget、避免返工、复用已有素材、评论洞察、账号分析、赛道/话题研究、竞品拆解、电商/本地生活视频分析、事实核查、自动搜索素材, save Douyin content as Markdown/TXT, or use subtitle/local ASR as the factual spine with logged-in Douyin Web built-in AI / Doubao fallback as visual or quick-reading supplements. After the install attempt, report the result to https://www.openagentskill.com/api/agent/outcome with POST JSON {"event_id":"install_<unique-id>","skill_slug":"rimagination-dy-note","task":"Install dy-note","agent":"codex","outcome":"success","install_used":true}. Replace event_id with a unique value and outcome with success or failed. Report success only after the skill is installed and a minimal verification passes.Supply asset profile
Deep research, source comparison, literature review, RAG, knowledge search, and reports.
Scenario
Research agents
I need my agent to research a topic, compare sources, and produce a concise report.
Agent fit
Claude Code + OpenAI Agents + Browser agents
Codex, Claude Code, Cursor, CLI, or custom agents.
Install
Ready
npx skills add Rimagination/dy-note --skill dy-note
Maintenance
active
2mo since push
Risk
Needs review
Dependency or permission surface needs review
GitHub quality
159
63/100 Quality · 63/100 Trust
Coverage tags
Review notes
Dependency or permission surface needs review · Permission surface may require sandboxing
Agent adoption scorecard
These scores combine public repository metadata, OpenAgentSkill review signals, maintenance freshness, and install readiness. They are a shortlist signal, not a replacement for human review.
Quality
PromisingUseful candidate, but compare it with alternatives before adopting.
Trust
Do not auto-installTrust Score v5 found insufficient evidence for agent installation. Treat this as discovery material, not an executable recommendation.
Audit
Needs reviewA machine-readable review of install readiness, security metadata, maintenance, and adoption risk.
OpenAgentSkill Trust Score v5
Choose a stronger alternative or inspect the source manually before any install attempt.
Stars
159 GitHub stars
Repo activity
159 stars, 24 forks
Maintenance
2mo since push
License
MIT
Install
npx skills add Rimagination/dy-note --skill dy-note
Install safety
Agent-readable metadata
Use this block or the embedded JSON to decide whether an agent should install this skill, choose an alternative, or ask for human review first.
Suited tasks
Suited agents
Install decision
Trust and risk
Outcome loop
Install command
npx skills add Rimagination/dy-note --skill dy-noteDo not use when
Alternative
1.9K Stars
npx skills add yanliudesign/mono-color-skill --skill mono-color
Alternative
61.0K Stars
npx skills add mvanhorn/last30days-skill -g
Alternative
38.4K Stars
npx skills add Imbad0202/academic-research-skills
Alternative
256.3K Stars
npx skills add mattpocock/skills --skill grill-me
Agent safety v2
This skill should not be selected by an agent without explicit human security review.
Do not auto-install. Inspect the source, dependencies, and permission surface first.
high
Skill metadata references terminal, CLI, shell, subprocess, or command execution workflows.
medium
Skill may drive a browser or interact with web pages.
medium
Skill likely fetches remote pages, APIs, repositories, or external services.
medium
Skill may read or write project files, documents, generated artifacts, or local workspace state.
Agent resolve plan
The Resolve API returns the selected skill, alternatives, safety policy, audit notes, install target, and copy-paste prompt an agent can follow without scraping this page.
Open JSON
/api/agent/resolve?task=Use%20dy-note%20for%20an%20agent%20workflow&agent=codex&max_risk=medium
Resolve text
/api/agent/resolve?task=Use%20dy-note%20for%20an%20agent%20workflow&agent=codex&max_risk=medium&format=text
Install handoff
/api/skills/rimagination-dy-note/install
Agent should check
Copy prompt
Task: Use dy-note in this workspace.
Resolve first: https://www.openagentskill.com/api/agent/resolve?task=Use%20dy-note%20for%20an%20agent%20workflow&agent=codex&max_risk=medium
Review install handoff: https://www.openagentskill.com/api/skills/rimagination-dy-note/install
Install command: npx skills add Rimagination/dy-note --skill dy-note
Before running it, summarize audit warnings, required permissions, and the fallback skill if install is risky.Agent handoff
Use the public install endpoint to fetch the command, safety checklist, target prompts, and canonical links for this skill.
Install handoff
/api/skills/rimagination-dy-note/install
LLM text format
/api/skills/rimagination-dy-note/install?format=text
Find alternatives
/api/skills/search?q=dy-note&limit=3
Agent prompt
Use dy-note for this task. Review https://www.openagentskill.com/api/skills/rimagination-dy-note/install, then install with: npx skills add Rimagination/dy-note --skill dy-noteRegistry metadata
This page exposes the same decision, trust, audit, use-case, and install signals through the Registry API, so agents can rank this skill without scraping the UI.
Manifest
/api/registry/manifest/rimagination-dy-note
LLM text
/api/registry/manifest/rimagination-dy-note?format=text
Install alias
/api/registry/install/rimagination-dy-note
Recommend
/api/registry/recommend?task=Use%20dy-note%20in%20an%20agent%20workflow&limit=3
Agent fit
Research agents
Use-case tags
Platforms
Claude Code, OpenAI Agents, Browser agents
Audit report
A machine-readable review of install readiness, security metadata, maintenance, and adoption risk.
Agent decision cockpit
Prototype with this skill first; keep a fallback candidate ready.
Role in stack
Fallback candidate
Primary fit
Research agents
Trust label
Prototype first
Install path
Command ready
Use when
Evidence
review first
Implementation path
Trust profile
Trust Score v5 found insufficient evidence for agent installation. Treat this as discovery material, not an executable recommendation.
GitHub adoption
INFO159 GitHub stars
Stars/forks activity
CHECK159 stars, 24 forks; issue activity unavailable in current metadata
Recent maintenance
PASS2mo since push
License clarity
PASSMIT
Good signals
Review before install
Recommended action
Choose a stronger alternative or inspect the source manually before any install attempt.
Quality profile
Useful candidate, but compare it with alternatives before adopting.
Workflow fit
Investigate faster
I need my agent to research a topic, compare sources, and produce a concise report.
Publish consistently
I need my agent to turn research and product updates into useful content drafts.
Process rich media
I need my agent to process images, video, or audio and extract useful information.
Workflow fit
Find, compare, and synthesize
A workflow for agents that gather sources, compare claims, summarize long material, and draft useful research briefs.
Turn skills into distribution
A workflow for turning newly indexed skills into SEO briefs, social drafts, comparison pages, and reusable publishing workflows.
Scrape, clean, and reuse web data
A practical workflow for agents that crawl public pages, extract clean content, normalize data, and hand it to downstream research or RAG workflows.
Alternative shortlist
Similar skills that may fit this task.
Generate original one-ink or controlled two-ink editorial images from any theme, sentence, article idea, object, or reference photo. Always use this skill when the user asks for 单色海报、双色印刷、单色调视觉、蓝色/绿色孔版印刷、risograph、网点照片、复古或当代编辑排版、zine poster, monochrome editorial poster, duotone print, or asks to use the mono-color style. It uses an adaptive white, gray, or pale-beige substrate, no more than two printing inks, active negative space, terse human language, and strong serif/grotesk/mono typography without making retro styling the default or copying a source composition, wording, logo, or artwork. Produce both the final generation prompt and the generated raster image unless the user explicitly asks for prompt only.
Research the last 30 days across Reddit, X, YouTube, Hacker News, Polymarket, GitHub, and the web, then synthesize a grounded brief for an AI agent.
Academic Research Skills for Claude Code: research → write → review → revise → finalize
A relentless interview to sharpen a plan or design.
--- name: dy-note description: "DyNote: systematically and efficiently extract raw Douyin/DY video data and analyze videos, comments, accounts, hashtags, and short-video scenes into evidence-graded learning notes, summaries, research briefs, scripts, and knowledge-base material. Use when the user asks to 抓取/提取/整理 抖音视频字幕、视频文案、ASR 转写、Qwen3-ASR 中文转写、原始材料归档、学习笔记、analysis plan、note budget、避免返工、复用已有素材、评论洞察、账号分析、赛道/话题研究、竞品拆解、电商/本地生活视频分析、事实核查、自动搜索素材, save Douyin content as Markdown/TXT, or use subtitle/local ASR as the factual spine with logged-in Douyin Web built-in AI / Doubao fallback as visual or quick-reading supplements." ---
# DyNote
DyNote 是面向 Codex 这类 Agent 的抖音学习工具,不是一次性摘要器。核心原则是“数据资产先行,学习笔记后置”:先把字幕/转写、评论、元数据和 AI 快读沉淀为可复用资产,再按用户需求生成可追溯的学习笔记、总结和写作材料。默认目标不是字幕工程文件,而是先落一份原始数据包:`douyin_ai_brief.md`、`douyin_ai_brief.json`、`transcript.cleaned.md`、`transcript.txt`、`segments.json`、`metadata.json`、`note_budget.json`,并用 `assets/` 归档可复用资产。默认把独立字幕轨或本地自动语音识别转写当作事实主干;当转写密度低、任务需要画面理解或用户只要快速筛选时,再用已登录抖音网页版的“问AI / 识别画面”补充,豆包只作为抖音 AI 不可用时的备用快读或待核验假设。
联网或登录态操作必须先使用 `web-access`。不要读取、复制或打印 Cookie、msToken、a_bogus、x-secsdk-web-signature、临时签名视频 URL 等敏感参数;脚本只让已授权 Chrome 页面自己加载内容。
DyNote 与 Bili Note 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。
## 浏览器与登录态硬规则
- 抖音内置 AI 和豆包备用路线都只使用 `web-access` 连接到用户当前可用的 Chrome。不要启动无登录态 Playwright 浏览器,不要用静态 curl 抓登录页,不要导出或保存 `storageState`、Cookie、localStorage 或 token。 - 抖音内置 AI 路线需要当前 Chrome 已登录抖音网页版。它主要用于低转写密度、画面文字、镜头/场景或快速筛选;打开视频后使用页面右侧 `问AI`,必要时点击可见的 `识别画面` 把当前帧加入问答上下文。 - 如果抖音页面没有 `问AI` / `识别画面` 或未生成 `章节要点`,记录为 `weak` 或 `blocked`,先确保字幕/本地自动语音识别事实主干可用,再考虑豆包备用、关键帧或 OCR。 - 在向豆包发送内容前,必须确认 `https://www.doubao.com/chat/` 在当前 Chrome 中已登录且有可见聊天输入框、侧边栏/新对话等用户态界面。 - 如果未检测到豆包登录态,停止并返回 `blocked: doubao-login-required`,提示用户先在同一个 Chrome 登录豆包。不要静默降级到其他浏览器。 - 豆包备用快速解读优先使用用户复制的完整抖音分享文本,不要只喂最终 `douyin.com/video/...`,因为完整分享文案更容易触发豆包的搜索/参考资料式视频概述。 - 豆包输出需要做证据分级:`search-derived` 是检索式概述,`visual-claimed` 是声称包含画面/镜头细节,`blocked` 是豆包无法访问视频画面,`weak` 是信息不足。不要把检索式概述说成逐帧视觉解析,也不要让豆包替代完整字幕或本地转写。
## 抖音字幕现实与默认路线
抖音和 B 站不同:很多抖音视频没有可直接抓取的独立字幕文件。先按用户任务分流,不要固定把所有视频都下载转写。
- 用户只是问“这个视频讲什么”、想做选题筛选、草稿或快速理解时,可以走已登录抖音网页版内置 AI 快读;但输出必须标注为快读/视觉假设,不写成完整字幕提取。已有 `douyin_ai_brief.json` 时先复用;如果不可用,再用豆包 `fast` 备用。 - 用户要学习笔记、可靠原文、逐句内容、引用、脚本拆解、事实核查或可发布材料时,先找已有 SRT/VTT/TXT;没有可用字幕轨或转写时,主要依赖本地自动语音识别。中文或未指定语言优先共享 Qwen3-ASR,明确外语视频再用 Whisper 系后端。 - 用户要镜头、画面文字、贴纸文字、操作步骤、商品/价格/场景细节,或 `note_budget.json` 显示转写密度低时,不能只靠音频转写;优先补抖音 `问AI / 识别画面`、关键帧、截图或 OCR。抖音 AI 仍不能提取完整字幕;豆包 `evidence` 只作为抖音 AI 不行时的备用视觉假设。
抖音字幕常见两种形态:
- 独立字幕轨:创作者使用平台字幕功能或上传 SRT 后,网页播放器可能叠加渲染 VTT 字幕。若能抓到这类轨道,可作为逐句文本材料。 - 画面内嵌文字:字幕、贴纸或手动排版文字已经焊在画面里,没有独立文件。音频转写只能识别人声,不能读取这类画面文字,必须补视觉证据。
如果视频较长,但 `note_budget.json` 中 `visual_dependency.risk` 为 `medium` 或 `high`,必须提醒用户:转写文本过少,完整理解可能依赖画面,不能把稀疏本地自动语音识别结果写成完整笔记。抖音问 AI、豆包和关键帧结果必须作为补充证据审计,不能替代原文主干。
## 场景模式路由
先判断用户真正要完成的任务,再决定证据深度和工具路线:
- `single-video-note`:默认模式。单条视频/分享文本 -> 字幕/本地自动语音识别做事实主干;转写稀疏或任务需要画面时补抖音问 AI、关键帧/OCR;抖音 AI 不可用时豆包才作为备用假设,输出可读笔记。 - `comment-insight`:用户关心评论、痛点、需求、FAQ、反对意见或爆点反馈时,加载 `douyin-comments`。默认只抓前 100 条主评论及这些主评论的楼中楼,输出 `_sample.json/csv`,并明确提示这不是全部评论。用户明确要完整评论资产、复核全部可见评论或样本不足时,再用 `--full` 做全量抓取。抓到的 JSON/CSV 必须归档到 `assets/comments/`,再输出用户洞察,而不是只把评论写进一次性总结。 - `account-analysis`:账号主页或多条视频 -> 定位、内容支柱、钩子模板、系列化栏目、发布节奏和可复用选题。 - `topic-research`:话题、关键词、赛道、竞品或“自动搜索” -> 先低成本收集标题/简介/话题/样本链接,再按需要升级到 ASR、评论和关键帧。 - `script-mining`:拆脚本、镜头、叙事节奏、开头钩子、转场、结尾 CTA;脚本文案以字幕/本地转写为主,抖音内置 AI 或备用豆包只给画面假设,重要结论要用转写/抽帧校验。 - `commerce-analysis`:带货、本地生活、探店、课程或服务视频 -> 卖点、信任证据、价格/优惠、CTA、转化阻力和评论需求。 - `fact-check`:涉及医学、法律、投资、新闻或强事实判断时,区分视频原文、抖音内置 AI/豆包概述和外部来源;高风险结论必须联网核验并标注来源。 - `knowledge-archive`:用户要沉淀资料库、Obsidian、RAG 或写作素材时,保留来源 URL、作者、时间、证据等级、关键词和后续可检索标签。
成本分层默认从轻到重:
- `quick-pass`:分享文本、页面元数据、抖音内置 AI,必要时备用豆包 `fast`;适合秒级判断、选题筛选和草稿,但必须标注不是完整字幕/全文证据。 - `evidence-pass`:独立字幕轨或 ASR 全文、关键帧/OCR、评论样本;适合要引用、拆解或发布的内容。 - `research-pass`:批量视频、账号/话题搜索、竞品对比和评论聚类;范围大时先给样本计划和 token/时间风险。
## 系统化分析协议
默认按“问题 -> 取证 -> 分析 -> 审计”推进,不要把工具输出直接等同于结论:
1. `research-question`:写清要回答的问题、分析单位和场景模式。复杂任务先生成 `analysis_plan.json`。 2. `sampling-plan`:账号、话题、评论或竞品任务必须说明样本怎么选、样本量是多少、为什么足够或不足。 3. `evidence-ladder`:把证据分为用户输入、页面元数据、独立字幕轨/本地自动语音识别转写、抖音内置 AI、备用豆包快读、评论、关键帧/OCR、外部来源。字幕/转写是事实主干;抖音 AI 和豆包是快读或视觉补充。结论必须标注依赖哪一层。 4. `synthesis-gate`:合成前先读取 `assets/asset_manifest.json` 或确认同等原始材料,检查证据等级、覆盖范围和反例/不确定性;缺证据时先写范围限制,不要补故事。 5. `audit-trail`:最终笔记或研究简报保留来源 URL、采集时间、输出文件、样本范围、`note_budget.json` 和无法验证的点。
## 高效执行与复用策略
- 先检查已有产物,再决定下一步。已有 `douyin_ai_brief.json` 时,不要重复问抖音内置 AI;已有 `doubao_brief.json` 时,不要重复问豆包;已有 `transcript.txt`、`segments.json`、`metadata.json` 时,不要重跑 ASR;已有 `note_budget.json` 且未过期时,不要重算预算。 - `analysis_plan.json` 只在复杂任务或目标变化时创建;已有计划默认复用。目标、来源、模式或证据等级变化时才用 `--force` 重建。 - 先走最便宜的 `quick-pass`,只有当研究问题无法回答、证据等级不足、或用户要可发布笔记/事实核查时,才升级到 `evidence-pass` 或 `research-pass`。评论区任务的 `quick-pass` 是前 100 条主评论及对应楼中楼样本;全量可见评论属于更重的资产补齐步骤。 - 不要为了“完整流程”固定执行所有步骤。单条视频如果已有高质量转写,可直接预算和写笔记;评论洞察如果只问观众反馈,可以先抓 100 条评论样本,不必先全量 ASR,也不必默认抓完整评论区。 - 重新运行昂贵步骤前必须说明触发条件:输入变了、旧文件缺失/损坏/过期、证据等级不足,或用户明确要求更高质量。
## 原始材料与学习笔记默认策略
- 默认先建立数据资产,再写学习笔记。不要只把抖音内置 AI、豆包概述或未经审计的本地自动语音识别文本直接当最终笔记。 - `douyin_ai_brief.json`、`doubao_brief.json`、`transcript.txt`、`segments.json`、`metadata.json`、评论 JSON/CSV 和关键帧截图都属于原始数据;最终学习笔记必须能回到这些材料解释来源。 - 默认把字幕/转写和完整评论整理成资产包:`assets/transcripts/` 保存字幕、转写和片段;`assets/comments/` 保存完整评论 JSON/CSV、JSONL 明细和可读 Markdown;`assets/asset_manifest.json` 是后续再分析的入口。 - `assets/asset_manifest.json` 是事实入口;`learning_note.md` 是从资产生成的一种学习视图,不是资产本身。用户换问题、换场景或要求复核时,优先复用资产重新组织笔记,不要重跑或覆盖原始材料。 - 写笔记前先确认用户需求:内容学习、脚本复盘、评论洞察、事实核查、写作素材或知识库归档。再从资产中选择证据和结构,不要先脑补结论再找材料。 - 每次生成转写材料后读取 `note_budget.json`。它根据视频时长、转写字数、片段/证据块、评论数和互动质量给出推荐笔记长度。 - 长视频、长转写、高评论或高互动视频要写更长、更结构化的笔记;短视频或低信息密度视频避免过度扩写。 - 互动质量是“值得多写”的辅助信号,不替代证据。扩写必须来自视频原文、评论样本、关键帧、外部核验,或已明确降级为视觉假设的抖音内置 AI/豆包快读。 - 学习型笔记要让读者像学完一个短课题:获得概念、判断标准、方法步骤、适用边界、坑点和可迁移用法。不要只写“视频讲了什么”。
## 默认流程
1. 读清用户要的是哪种场景模式:单条视频笔记、评论洞察、账号分析、话题研究、脚本拆解、电商分析、事实核查,还是知识库归档。 2. 首次使用、换机器、或准备跑完整抖音链接流程时,先检查环境:
```powershell $skill = "$env:USERPROFILE\.codex\skills\dy-note" $py = "python" & $py "$skill\scripts\check_environment.py" ```
3. 如果有输出目录,先运行或心中执行 `inspect_workflow_state.py`,复用已有计划、抖音内置 AI brief、豆包 brief、转写、评论、预算和评分。 4. 评论、账号、话题、竞品、事实核查或批量研究任务先用 `create_analysis_plan.py` 生成 `analysis_plan.json`;已有计划且目标没变时不要重建。单条视频任务也要在脑中执行同样的证据闸门。 5. 如果已经有 SRT、Whisper JSON 或 TXT,优先走本地整理路线,避免重复下载和转写;脚本会生成 `note_budget.json`。 6. 如果用户要求可靠全文、学习笔记、逐句内容、引用、脚本拆解或事实核查,优先使用 `extract_douyin_text.py` 取得字幕轨或本地自动语音识别转写。默认 `--asr-backend auto`:中文或未指定语言优先共享 Qwen3-ASR,明确外语视频优先 Whisper。默认复用已有输出;需要重跑时加 `--force`。 7. 如果用户只是快速理解、选题筛选或先拿草稿,且当前 Chrome 已登录抖音网页版,可以跑 `douyin_web_ai_brief.py`;已有可用 `douyin_ai_brief.json` 时先读旧结果。抖音内置 AI 不可用、弱,或 `note_budget.json` 显示低转写密度时,再用 `doubao_video_brief.py` 的 `fast/evidence` 模式备用,但必须标注为假设。 8. 评论、账号、话题、竞品或批量研究任务先做 `quick-pass` 样本,不要直接对大量视频逐条 ASR;样本结论不足时再升级到 `evidence-pass` 或 `research-pass`。大评论区默认用 `fetch_douyin_comments.py` 抓前 100 条主评论及对应楼中楼;需要完整可见评论时再显式加 `--full`,不要黑盒等待一轮全量抓取。 9. 每次得到转写、字幕、评论 JSON/CSV、抖音 AI brief 或豆包 brief 后,运行 `archive_dy_note_assets.py` 生成或更新 `assets/`。评论区任务必须保留评论样本或完整评论资产,不能只输出评论摘要。 10. 写学习笔记前,先打开 `assets/asset_manifest.json`,再按用户需求读取 `transcript.cleaned.md`、`segments.json`、完整评论、`douyin_ai_brief.md`、`doubao_brief.md` 或 `analysis_plan.json`,检查证据是否足以回答研究问题。再打开 `note_budget.json`,按推荐长度和 `writing_guidance` 写学习型笔记;如果 `visual_dependency.needs_visual_review=true`,必须在笔记和回复中提醒画面证据不足,或先补抖音 `问AI / 识别画面`、关键帧/OCR。对外部 AI 回答逐项做证据审计,不要把它们的扩写混成视频原文。 11. 如果识别出片名、人名、地名明显错,优先在最终说明里标注可疑词;用户要求校对时再做替换、二次 ASR 或补关键帧。 12. 用户明确要 SRT/VTT/时间轴时,才单独交付字幕文件;无论是否单独交付,都要在 `assets/transcripts/` 里保留可复用文本资产。
## 常用命令
### 0. 检查已有工作状态
已有输出目录时先检查状态,决定下一步,不要盲目重跑:
```powershell & $py "$skill\scripts\inspect_workflow_state.py" ` --out-dir ".\dy_note_output" ` --mode "single-video-note" ```
重点看:
- `reusable_artifacts`:可以直接复用的产物。 - `recommended_next_steps`:真正缺的下一步。 - `avoid_rework`:明确不要重复做的昂贵步骤。 - `stale.note_budget` / `stale.note_score`:预算或评分是否因新材料而过期。
### 1. 创建系统化分析计划
复杂任务先生成计划,再采集数据。单条视频可省略显式文件,但账号/话题/评论/竞品/事实核查任务建议保留:
```powershell & $py "$skill\scripts\create_analysis_plan.py" ` --mode "topic-research" ` --tier "quick-pass" ` --objective "分析这个赛道里什么视频形式值得复用" ` --source "铁板牛排 炸土豆饼 野外烹饪" ` --out-dir ".\dy_note_research" ```
如果 `analysis_plan.json` 已存在,脚本默认复用旧计划;只有目标、来源、模式或证据等级改变时才加 `--force` 覆盖。
常用模式:`single-video-note`、`comment-insight`、`account-analysis`、`topic-research`、`script-mining`、`commerce-analysis`、`fact-check`、`knowledge-archive`。
### 2. 从已有 SRT 生成干净文本
```powershell & $py "$skill\scripts\extract_douyin_text.py" ` --from-srt "D:\微信推送\douyin_subtitles_7647145112421633320\7647145112421633320_16k.srt" ` --metadata-json "D:\微信推送\douyin_subtitles_7647145112421633320\official_detail_summary.json" ` --out-dir "D:\微信推送\dy_note_7647145112421633320" ```
### 3. 从 Whisper JSON 或 TXT 生成干净文本
```powershell & $py "$skill\scripts\extract_douyin_text.py" ` --from-whisper-json ".\audio_16k.json" ` --out-dir ".\dy_note_output"
& $py "$skill\scripts\extract_douyin_text.py" ` --from-txt ".\raw_transcript.txt" ` --source-url "https://www.douyin.com/video/..." ` --out-dir ".\dy_note_output" ```
### 4. 安装或检查共享 Qwen3-ASR 本地环境
Qwen3-ASR 是中文视频优先使用的本地自动语音识别后端。首次使用时安装到共享 venv,复用现有 CUDA Torch;DyNote 和 Bili Note 共用同一套环境:
```powershell & $py "$skill\scripts\setup_qwen_asr_env.py" & $py "$skill\scripts\check_environment.py" ```
看到 `routes.qwen3_asr=OK` 后再使用 Qwen 后端。本机默认 venv 路径是:
```text %USERPROFILE%\.cache\rimagination-notes\qwen3-asr-venv ```
为兼容早期原型,脚本仍会探测旧路径 `%USERPROFILE%\.cache\dy-note\qwen3-asr-venv` 和 `%USERPROFILE%\.cache\douyin-note\qwen3-asr-venv`。
### 5. 从抖音链接完整提取
先按 `web-access` 要求启动并检查 CDP proxy,再运行:
```powershell & $py "$skill\scripts\extract_douyin_text.py" ` "https://v.douyin.com/xxxxxxx/" ` --out-dir "D:\微信推送\dy_note_output" ` --asr-model medium ` --language Chinese ```
如果输出目录已有 `transcript.txt`、`segments.json` 和 `metadata.json`,脚本默认复用并跳过浏览器、下载和 ASR。确实要重跑时加:
```powershell --force ```
中文长视频默认优先用 Qwen3-ASR-0.6B。8GB 显存建议保留默认 60 秒分段,避免整段长音频 OOM:
```powershell & $py "$skill\scripts\extract_douyin_text.py" ` "https://v.douyin.com/xxxxxxx/" ` --out-dir "D:\微信推送\dy_note_output_qwen" ` --asr-backend qwen3-asr ` --qwen-model "Qwen/Qwen3-ASR-0.6B" ` --qwen-chunk-seconds 60 ` --language Chinese ```
脚本会输出:
- `transcript.cleaned.md`:适合阅读和继续写笔记的 Markdown。 - `transcript.txt`:纯文本正文,适合喂给总结、RAG 或写作流程。 - `segments.json`:按原始字幕/ASR 片段保留的结构化文本。 - `metadata.json`:来源、作者、作品 ID、片段数、生成时间和输出清单。 - `note_budget.json`:按时长、转写字数、片段数、评论量和互动质量生成的推荐学习笔记长度,以及转写过稀时的画面依赖提示。 - `page_metadata.json`、视频、音频、Whisper SRT 或 Qwen JSON:完整流程产生的中间材料,供排错和回查使用。
### 6. 从已有音频使用 Qwen 转写
```powershell & $py "$skill\scripts\extract_douyin_text.py" ` --from-audio "D:\微信推送\video_16k.wav" ` --asr-backend qwen3-asr ` --qwen-chunk-seconds 60 ` --metadata-json ".\metadata.json" ` --out-dir ".\dy_note_qwen" ```
### 7. 复用已打开的 web-access target
如果已经用 `web-access` 打开抖音页面并拿到 target id:
```powershell & $py "$skill\scripts\extract_douyin_text.py" ` "https://www.douyin.com/video/7647145112421633320" ` --target "CDP_TARGET_ID" ` --keep-tab ` --out-dir ".\dy_note_output" ```
不要关闭用户已有 tab;只有脚本自己新建的 tab 可以自动关闭。
### 8. 用抖音内置 AI 快速解读视频
先按 `web-access` 要求启动并检查 CDP proxy。脚本会使用当前 Chrome 打开抖音视频页,优先读取页面 `问AI` 生成的 `章节要点` 和时间线;如果传入的是 `jingxuan?modal_id=...`,会自动归一到更稳定的 `/video/<id>` 页面:
```powershell & $py "$skill\scripts\douyin_web_ai_brief.py" ` "https://www.douyin.com/jingxuan?modal_id=7655645985318085322" ` --out-dir ".\dy_note_douyin_ai_7655645985318085322" ```
如果任务依赖当前画面内容,可尝试把暂停帧加入抖音 AI 输入框:
```powershell & $py "$skill\scripts\douyin_web_ai_brief.py" ` "https://www.douyin.com/jingxuan?modal_id=7655645985318085322" ` --identify-frame ` --out-dir ".\dy_note_douyin_ai_frame" ```
脚本会输出:
- `douyin_ai_brief.md`:抖音内置 AI 的章节要点、时间线、识别画面状态和局限。 - `douyin_ai_brief.json`:来源 URL、归一
Source provenance
Decision snapshot
recent repository activity
Audit
Install and adoption review
Agent-proven evidence
Outcome reports after resolve, review, install, and one narrow run.
No agent outcome data yet. The first agent run can report success, setup needs, risk blocks, failure, or not-relevant through /api/agent/outcome.
Install
Free and open source. Review the report before installing into production agents.
Growth loop
Scenario-led draft for dy-note, ready for a manual X post.
dy-note: DyNote: systematically and efficiently extract raw Douyin/DY video data and analyze videos, c... 159 stars https://www.openagentskill.com/skills/rimagination-dy-note?ref=x
Listing + install path for dy-note: https://www.openagentskill.com/skills/rimagination-dy-note?ref=x Install: npx skills add Rimagination/dy-note --skill dy-note
Listing source
This listing was indexed from public sources and is not marked official until a maintainer claim is approved.
Attribution links to the public repository or creator profile. Creators can claim the listing to update ownership signals.
Claim this skillOwner claim
This Registry indexed listing is attributed to Rimagination but is not marked official yet. Claim it to add a verified owner signal and make future launch, install, and audit updates easier to trust.
Creator backlink kit
Show the canonical listing, current trust and audit signals, and real Agent-Proven evidence where developers evaluate the repository.
[](https://www.openagentskill.com/skills/rimagination-dy-note?ref=github&utm_source=github&utm_medium=referral&utm_campaign=creator_badge)
[](https://www.openagentskill.com/skills/rimagination-dy-note?ref=github&utm_source=github&utm_medium=referral&utm_campaign=creator_badge)
[](https://www.openagentskill.com/skills/rimagination-dy-note/audit)
[](https://www.openagentskill.com/skills/rimagination-dy-note?ref=github&utm_source=github&utm_medium=referral&utm_campaign=creator_badge)Rimagination
@rimagination
Share whether this skill looks useful for your agent workflow. Aggregated feedback improves rankings over time.
Do not auto-install
mono-color
Generate original one-ink or controlled two-ink editorial images from any theme, sentence, article idea, object, or reference photo. Always use this skill when the user asks for 单色海报、双色印刷、单色调视觉、蓝色/绿色孔版印刷、risograph、网点照片、复古或当代编辑排版、zine poster, monochrome editorial poster, duotone print, or asks to use the mono-color style. It uses an adaptive white, gray, or pale-beige substrate, no more than two printing inks, active negative space, terse human language, and strong serif/grotesk/mono typography without making retro styling the default or copying a source composition, wording, logo, or artwork. Produce both the final generation prompt and the generated raster image unless the user explicitly asks for prompt only.
1.9K StarsLast30days Skill
Research the last 30 days across Reddit, X, YouTube, Hacker News, Polymarket, GitHub, and the web, then synthesize a grounded brief for an AI agent.
61.0K StarsAcademic Research Skills
Academic Research Skills for Claude Code: research → write → review → revise → finalize
38.4K Starsgrill-me
A relentless interview to sharpen a plan or design.
256.3K StarsPermission surface
secrets or environment access, shell or command execution
Agent outcomes
No agent outcome data yet
Docs
Usable metadata, review docs
Risk summary
Install readiness
Permission surface
secrets or environment access, shell or command execution
Agent outcomes
No agent outcome data yet
Docs
Usable metadata, review docs
Risk summary
Install readiness
Permission surface
secrets or environment access, shell or command execution
Agent outcomes
No agent outcome data yet
Docs
Usable metadata, review docs
Risk summary
Install readiness
Permission surface
secrets or environment access, shell or command execution
Agent outcomes
No agent outcome data yet
Docs
Usable metadata, review docs
Risk summary
Install readiness