Skip to content

anbeime/skill / skills/anything2explainer/anything2explainer

anything2explainer

Unverified

给一个主题,产出一条黑底 MG 风格(幕底可选星点或点阵波)、有配音字幕章节进度条的科普讲解视频(中文或英文;Remotion 代码动画;时长由用户定,常用 3–5 分钟)。内含可编译模板、图元库、配音/分镜/渲染工具、风格与动效规范、多 agent 分工协议与 QC 判据,以及一条完整样片(《RAG 与知识库》)作为质量标尺。Turn any topic into a narrated motion-graphics explainer video in Chinese or English, on a black canvas with a star-field or dot-field backdrop, TTS voiceover, subtitles and a chapter progress bar, every frame drawn in code with Remotion. Use when the user asks for an explainer, educational or science-communication video about a topic, or wants an article or document turned into a video.

FileKindSize
.gitignoreother397 B
CITATION.cffother1.1 KB
LICENSEother4.9 KB
README.mdmarkdown17.2 KB
README_ZH.mdmarkdown16.0 KB
SKILL.mdskill md20.9 KB
examples/contrast/01_hero_size_bad_f0128.jpgbinary44.2 KB
examples/contrast/01_hero_size_good_f0138.jpgbinary50.0 KB
examples/contrast/02_hero_reveal_bad_f1729.jpgbinary49.3 KB
examples/contrast/02_hero_reveal_good_f1105.jpgbinary43.6 KB
examples/contrast/03_big_number_good_f7258.jpgbinary47.2 KB
examples/contrast/03_sparse_axis_bad_f1900.jpgbinary35.3 KB
examples/contrast/04_formula_bad_f3033.jpgbinary55.3 KB
examples/contrast/04_result_number_good_f2613.jpgbinary82.2 KB
examples/contrast/05_clean_bg_good_f1237.jpgbinary61.3 KB
examples/contrast/05_clutter_bg_bad_f1651.jpgbinary36.5 KB
examples/contrast/06_ending_bad_f7540.jpgbinary48.0 KB
examples/contrast/06_ending_good_f8100.jpgbinary57.2 KB
examples/contrast/README.mdmarkdown1.4 KB
examples/contrast/contrast_sheet.jpgbinary398.9 KB
examples/rag/AGENT_RAG_BUILD_RULES.mdmarkdown9.4 KB
examples/rag/AGENT_RAG_QC_RULES.mdmarkdown2.9 KB
examples/rag/README.mdmarkdown1.0 KB
examples/rag/frames/overview_1.jpgbinary282.3 KB
examples/rag/frames/overview_2.jpgbinary331.9 KB
examples/rag/frames/overview_3.jpgbinary326.7 KB
examples/rag/frames/overview_4.jpgbinary374.1 KB
examples/rag/frames/overview_5.jpgbinary319.4 KB
examples/rag/frames/overview_6.jpgbinary285.3 KB
examples/rag/frames/ref_agentic_loop_f6900.jpgbinary71.6 KB
examples/rag/frames/ref_chapter_card_f1490.jpgbinary35.8 KB
examples/rag/frames/ref_chunk_size_f2200.jpgbinary86.5 KB
examples/rag/frames/ref_graphrag_f6600.jpgbinary71.6 KB
examples/rag/frames/ref_hnsw_f3260.jpgbinary85.3 KB
examples/rag/frames/ref_hybrid_rrf_f4640.jpgbinary95.8 KB
examples/rag/frames/ref_long_context_f7480.jpgbinary63.0 KB
examples/rag/frames/ref_lost_in_middle_f5250.jpgbinary89.0 KB
examples/rag/frames/ref_moat_f8100.jpgbinary71.4 KB
examples/rag/frames/ref_paper_f1300.jpgbinary88.4 KB
examples/rag/frames/ref_parse_f1760.jpgbinary74.0 KB
examples/rag/frames/ref_rag_reveal_f1090.jpgbinary45.1 KB
examples/rag/frames/ref_ragas_f6150.jpgbinary129.7 KB
examples/rag/frames/ref_rerank_funnel_f4900.jpgbinary80.1 KB
examples/rag/frames/ref_shortcomings_f0600.jpgbinary83.6 KB
examples/rag/frames/ref_title_f0058.jpgbinary48.9 KB
examples/rag/frames/ref_vector_space_f2990.jpgbinary93.2 KB
examples/rag/narration.txtmarkdown4.4 KB
examples/rag/qc/qc_v1_C1.mdmarkdown14.4 KB
examples/rag/qc/qc_v1_C2.mdmarkdown22.5 KB
examples/rag/qc/qc_v1_C3.mdmarkdown16.7 KB
examples/rag/qc/qc_v1_C4.mdmarkdown23.0 KB
examples/rag/qc/qc_v3_final.mdmarkdown12.7 KB
examples/rag/research.mdmarkdown47.7 KB
examples/rag/shots_src/G0/Overlay.tsxother10.5 KB
examples/rag/shots_src/G0/Preview.tsxother224 B
examples/rag/shots_src/G0/index.tsscript1.4 KB
examples/rag/shots_src/G0/util.tsscript167 B
examples/rag/shots_src/G1/BUILD_NOTES.mdmarkdown12.1 KB
examples/rag/shots_src/G1/Preview.tsxother344 B
examples/rag/shots_src/G1/SC01.tsxother448 B
examples/rag/shots_src/G1/SC02.tsxother4.6 KB
examples/rag/shots_src/G1/SC03.tsxother4.7 KB
examples/rag/shots_src/G1/SC04.tsxother4.9 KB
examples/rag/shots_src/G1/SC05.tsxother7.0 KB
examples/rag/shots_src/G1/index.tsscript664 B
examples/rag/shots_src/G1/layout.tsxother6.8 KB
examples/rag/shots_src/G2/BUILD_NOTES.mdmarkdown15.3 KB
examples/rag/shots_src/G2/Preview.tsxother344 B
examples/rag/shots_src/G2/SC06.tsxother4.0 KB
examples/rag/shots_src/G2/SC07.tsxother5.1 KB
examples/rag/shots_src/G2/SC08.tsxother6.1 KB
examples/rag/shots_src/G2/SC09.tsxother7.1 KB
examples/rag/shots_src/G2/SC10.tsxother5.6 KB
examples/rag/shots_src/G2/index.tsscript657 B
examples/rag/shots_src/G2/parts.tsxother10.7 KB
examples/rag/shots_src/G3/BUILD_NOTES.mdmarkdown13.2 KB
examples/rag/shots_src/G3/Preview.tsxother344 B
examples/rag/shots_src/G3/SC11.tsxother3.9 KB
examples/rag/shots_src/G3/SC12.tsxother3.0 KB
examples/rag/shots_src/G3/SC13.tsxother2.3 KB
examples/rag/shots_src/G3/SC14.tsxother5.1 KB
examples/rag/shots_src/G3/SC15.tsxother7.3 KB
examples/rag/shots_src/G3/g3ui.tsxother12.1 KB
examples/rag/shots_src/G3/index.tsscript660 B
examples/rag/shots_src/G4/BUILD_NOTES.mdmarkdown13.1 KB
examples/rag/shots_src/G4/Preview.tsxother344 B
examples/rag/shots_src/G4/SC16.tsxother5.1 KB
examples/rag/shots_src/G4/SC17.tsxother5.9 KB
examples/rag/shots_src/G4/SC18.tsxother9.8 KB
examples/rag/shots_src/G4/SC19.tsxother6.9 KB
examples/rag/shots_src/G4/SC20.tsxother8.3 KB
examples/rag/shots_src/G4/index.tsscript660 B
examples/rag/shots_src/G4/vspace.tsscript3.2 KB
examples/rag/shots_src/G5/BUILD_NOTES.mdmarkdown12.9 KB
examples/rag/shots_src/G5/Preview.tsxother344 B
examples/rag/shots_src/G5/SC21.tsxother5.3 KB
examples/rag/shots_src/G5/SC22.tsxother5.0 KB
examples/rag/shots_src/G5/SC23.tsxother5.3 KB
examples/rag/shots_src/G5/SC24.tsxother5.7 KB
examples/rag/shots_src/G5/SC25.tsxother8.5 KB
examples/rag/shots_src/G5/g5util.tsxother3.4 KB
examples/rag/shots_src/G5/index.tsscript660 B
examples/rag/shots_src/G5/vspace.tsxother3.7 KB
examples/rag/shots_src/G6/BUILD_NOTES.mdmarkdown12.8 KB
examples/rag/shots_src/G6/Preview.tsxother344 B
examples/rag/shots_src/G6/SC26.tsxother7.9 KB
examples/rag/shots_src/G6/SC27.tsxother3.8 KB
examples/rag/shots_src/G6/SC28.tsxother7.9 KB
examples/rag/shots_src/G6/SC29.tsxother5.8 KB
examples/rag/shots_src/G6/SC30.tsxother3.0 KB
examples/rag/shots_src/G6/index.tsscript881 B
examples/rag/shots_src/G6/ui.tsxother12.4 KB
examples/rag/shots_src/G7/BUILD_NOTES.mdmarkdown12.9 KB
examples/rag/shots_src/G7/Preview.tsxother344 B
examples/rag/shots_src/G7/SC31.tsxother1.2 KB
examples/rag/shots_src/G7/SC32.tsxother3.3 KB
examples/rag/shots_src/G7/SC33.tsxother5.4 KB
examples/rag/shots_src/G7/SC34.tsxother5.8 KB
examples/rag/shots_src/G7/SC35.tsxother2.7 KB
examples/rag/shots_src/G7/SC36.tsxother7.8 KB
examples/rag/shots_src/G7/SC37.tsxother10.4 KB
examples/rag/shots_src/G7/g7ui.tsxother9.5 KB
examples/rag/shots_src/G7/index.tsscript818 B
examples/rag/shots_src/G8/BUILD_NOTES.mdmarkdown14.1 KB
examples/rag/shots_src/G8/Preview.tsxother375 B
examples/rag/shots_src/G8/SC38.tsxother3.9 KB
examples/rag/shots_src/G8/SC39.tsxother2.7 KB
examples/rag/shots_src/G8/SC40.tsxother899 B
examples/rag/shots_src/G8/SC41.tsxother4.8 KB
examples/rag/shots_src/G8/SC42.tsxother3.1 KB
examples/rag/shots_src/G8/SC43.tsxother2.7 KB
examples/rag/shots_src/G8/SC44.tsxother4.5 KB
examples/rag/shots_src/G8/index.tsscript818 B
examples/rag/shots_src/G8/shared.tsxother21.0 KB
examples/rag/storyboard_src.mdmarkdown34.8 KB
examples/rag/timeline.mdmarkdown6.2 KB
examples/rag/ui_rag.tsxother19.2 KB
examples/rag/交付说明.mdmarkdown4.6 KB
examples/rag/分镜表.mdmarkdown32.7 KB
reference/agent-build-rules.mdmarkdown13.0 KB
reference/agent-qc-rules.mdmarkdown6.2 KB
reference/composition-and-light.mdmarkdown11.5 KB
reference/lessons.mdmarkdown32.1 KB
reference/motion-vocabulary.mdmarkdown10.1 KB
reference/narration-guidance.mdmarkdown7.0 KB
reference/narration-storyboard.mdmarkdown21.4 KB
reference/prompts.mdmarkdown8.2 KB
reference/research-brief.mdmarkdown3.4 KB
reference/style-guide.mdmarkdown10.5 KB
template/.gitignoreother90 B
template/package.jsonmanifest457 B
template/public/fonts/Audiowide-Regular.ttfbinary68.3 KB
template/public/fonts/Exo2-Italic.ttfbinary295.5 KB
template/public/fonts/LICENSE.mdmarkdown5.4 KB
template/public/fonts/NotoSansSC.ttfbinary16.9 MB
template/public/fonts/Orbitron[wght].ttfbinary37.7 KB
template/remotion.config.tsscript1.1 KB
template/script/narration.txtmarkdown591 B
template/scripts/frame_metrics.pyscript9.4 KB
template/scripts/motion_check.pyscript7.9 KB
template/scripts/new_project.shscript1.4 KB
template/scripts/preview.shscript1.1 KB
template/scripts/render.shscript989 B
template/scripts/render_storyboard.pyscript1.3 KB
template/scripts/selfcheck.pyscript6.2 KB
template/scripts/sheet.pyscript1.6 KB
template/scripts/still.shscript1.1 KB
template/scripts/test_render.shscript732 B
template/scripts/tts_build.pyscript28.5 KB
template/src/Main.tsxother2.5 KB
template/src/Root.tsxother1.8 KB
template/src/common/DotFieldBg.tsxother3.7 KB
template/src/common/Fog.tsxother1.5 KB
template/src/common/Footage.tsxother2.8 KB
template/src/common/Glitch.tsxother5.3 KB
template/src/common/ProgressBar.tsxother4.0 KB
template/src/common/StarField.tsxother5.6 KB
template/src/common/Subtitle.tsxother3.1 KB
template/src/common/easing.tsscript5.4 KB
template/src/common/index.tsscript462 B
template/src/common/lib.tsxother4.5 KB
template/src/common/subs.tsscript175 B
template/src/common/textfit.tsscript3.3 KB
template/src/common/timeline.tsscript355 B
template/src/common/types.tsscript870 B
template/src/config.tsscript4.0 KB
template/src/fx.tsxother16.8 KB
template/src/index.tsscript89 B
template/src/overlay/Overlay.tsxother13.7 KB
template/src/overlay/Preview.tsxother250 B
template/src/overlay/index.tsscript1.6 KB
template/src/shots/G1/Preview.tsxother356 B
template/src/shots/G1/index.tsscript478 B
template/src/shots/G2/Preview.tsxother356 B
template/src/shots/G2/index.tsscript478 B
template/src/shots/G3/Preview.tsxother356 B
template/src/shots/G3/index.tsscript478 B
template/src/shots/G4/Preview.tsxother356 B
template/src/shots/G4/index.tsscript478 B
template/src/shots/G5/Preview.tsxother356 B

examples/rag/timeline.md

View on GitHub
# 时间轴(zh-CN-YunxiNeural +8%,共 8246 帧 = 274.9s,1490 字,语速 5.96 字/s)

| 句 | 章 | 帧 from–to | 时长 | 文本(| 为字幕切分) |
|---|---|---|---|---|
| S01 | 1 | 86–165 | 2.7s | 大模型很聪明|但它有三个天生的短板 |
| S02 | 1 | 176–300 | 4.2s | 第一|它的知识有截止日期|训练之后发生的事|它一概不知 |
| S03 | 1 | 311–443 | 4.4s | 第二|它读不到你公司内网里的文档|也不知道你昨天刚写的会议纪要 |
| S04 | 1 | 454–610 | 5.2s | 第三|遇到不知道的问题|它不会说我不知道|而是一脸自信地编一个答案|这就是幻觉 |
| S05 | 1 | 621–741 | 4.0s | 怎么办|重新训练一遍|成本太高|而且明天数据又变了 |
| S06 | 1 | 752–868 | 3.9s | 于是工程师们想出了一个办法|让模型开卷考试 |
| S07 | 1 | 879–1010 | 4.4s | 考试前先把资料整理成一本手册|答题时先翻手册|再作答 |
| S08 | 1 | 1021–1127 | 3.6s | 这套方法就是今天的主角|RAG|检索增强生成 |
| S09 | 1 | 1138–1326 | 6.3s | 2020年|Facebook AI 的研究者在论文里|正式提出了这个名字|Retrieval-Augmented Generation |
| S10 | 1 | 1337–1478 | 4.7s | 它把检索和生成两个动作接在一起|回答不再只靠记忆|而是靠证据 |
| S11 | 2 | 1534–1652 | 4.0s | RAG 好不好用|一半取决于那本手册|也就是知识库是怎么建的 |
| S12 | 2 | 1663–1888 | 7.5s | 第一步解析|PDF 网页 表格 扫描件|要先统一变成干净的文本|还得保留标题段落和表格结构 |
| S13 | 2 | 1899–2052 | 5.1s | 第二步切块|模型一次读不完一整本手册|所以要把文档切成一段段的 chunk |
| S14 | 2 | 2063–2306 | 8.1s | 切多大是个学问|太大检索不精准|太小上下文被切碎|常见做法是几百个 token 一块|块与块之间留一点重叠 |
| S15 | 2 | 2317–2683 | 12.2s | 更讲究的做法|是按语义或标题层级来切|再给每个小块补一句它来自哪篇文档的哪一节|Anthropic 的实验显示|配合关键词检索|这一招能把检索失败率降低 49% |
| S16 | 2 | 2694–2858 | 5.5s | 第三步向量化|用 Embedding 模型|把每一块文字压成一串几百到几千维的数字 |
| S17 | 2 | 2869–3031 | 5.4s | 意思相近的文字|向量就靠得近|退款流程和怎么把钱要回来|在这个空间里是邻居 |
| S18 | 2 | 3042–3296 | 8.5s | 第四步建索引|上亿个向量不能一个个比|要用 HNSW 这类近似最近邻索引|把查找从线性扫描变成几十次跳转 |
| S19 | 2 | 3307–3480 | 5.8s | 存进向量数据库|再配上元数据和权限标签|谁能看什么|在这一步就要定下来 |
| S20 | 2 | 3491–3660 | 5.7s | 别忘了文档会变|知识库要能增量更新|否则模型很快又在讲过期的故事 |
| S21 | 3 | 3716–3819 | 3.5s | 手册建好了|用户提问的那一刻|真正的考试才开始 |
| S22 | 3 | 3830–4054 | 7.5s | 第一步理解问题|用户的原话往往含糊|可以让模型先改写|拆成几个子问题|或者先假想一个答案|再拿它去检索 |
| S23 | 3 | 4065–4223 | 5.3s | 第二步检索|把问题也变成向量|去索引里找最近的邻居|这是语义检索 |
| S24 | 3 | 4234–4412 | 6.0s | 但语义检索有盲区|型号 编号 专有名词这种精确匹配|它经常抓不住 |
| S25 | 3 | 4423–4674 | 8.4s | 所以工程上通常做混合检索|向量找相似的|BM25 关键词找精确的|两路结果用倒数排名融合成一份 |
| S26 | 3 | 4685–4924 | 8.0s | 第三步重排|粗召回几十条|再用 Cross-Encoder 重排器|把问题和每一段放在一起精细打分|只留最相关的几条 |
| S27 | 3 | 4935–5015 | 2.7s | 这一步很像推荐系统里的精排|贵|但准 |
| S28 | 3 | 5026–5311 | 9.5s | 第四步组装上下文|研究发现|模型对塞在长上下文中间的信息记得最差|这叫 Lost in the Middle|所以最重要的证据要放在开头和结尾 |
| S29 | 3 | 5322–5562 | 8.0s | 最后生成|把问题和证据一起交给大模型|并要求它只根据给定资料回答|答不出来就说不知道|每句结论都标注来源 |
| S30 | 3 | 5573–5694 | 4.1s | 一个有引用可追溯的回答|才是 RAG 真正的价值 |
| S31 | 4 | 5750–5819 | 2.3s | 系统搭起来了|怎么知道它好不好 |
| S32 | 4 | 5830–5944 | 3.8s | 评估要拆成两半|检索有没有找对|生成有没有说对 |
| S33 | 4 | 5955–6210 | 8.5s | 检索端看召回率 MRR 和 nDCG|生成端看忠实度和答案相关性|RAGAS 这类框架把它们做成了标准指标 |
| S34 | 4 | 6221–6404 | 6.1s | 一个常见的坑|回答很流畅|但证据是错的|所以忠实度永远比流畅度重要 |
| S35 | 4 | 6415–6448 | 1.1s | RAG 也在进化 |
| S36 | 4 | 6459–6707 | 8.3s | GraphRAG|把文档里的实体和关系抽成知识图谱|再对社区做摘要|专门回答这批文档整体讲了什么|这类全局问题 |
| S37 | 4 | 6718–6988 | 9.0s | Agentic RAG|让模型自己决定要不要检索|检索几轮|结果够不够|不够就换个问法再搜|今天的 Deep Research 类产品|本质上就是它 |
| S38 | 4 | 6999–7165 | 5.6s | 多模态 RAG|直接对图片和 PDF 页面做向量|图表 截图也能被检索到 |
| S39 | 4 | 7176–7299 | 4.1s | 有人问|模型上下文都百万 token 了|还需要 RAG 吗 |
| S40 | 4 | 7310–7501 | 6.4s | Anthropic 给过一条经验线|资料不到 20 万 token|大约 500 页|直接整本塞进上下文就够了 |
| S41 | 4 | 7512–7751 | 8.0s | 但一家公司的资料远不止 500 页|还有权限 时效和成本|RAG 负责找|长上下文负责读|两者是搭档不是对手 |
| S42 | 4 | 7762–7867 | 3.5s | 回到最初的比喻|RAG 让模型从闭卷变成开卷 |
| S43 | 4 | 7878–8050 | 5.8s | 但开卷考试想考高分|靠的从来不是翻书翻得快|而是那本手册整理得有多好 |
| S44 | 4 | 8061–8146 | 2.9s | 知识库|才是 RAG 真正的护城河 |

## 章节起始帧
- 第1章 为什么要RAG:f86
- 第2章 知识库构建:f1534
- 第3章 检索与生成:f3716
- 第4章 评估与进阶:f5750
reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis.Stars: 468kUpdated Claude CodeCodexVerifiedInstall
diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness.Stars: 321kUpdated Claude CodeCodexVerifiedInstall
dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependenciesStars: 321kUpdated Claude CodeCodexVerifiedInstall
executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is availableStars: 321kUpdated Claude CodeCodexVerifiedInstall