[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M6BY2W6MYA5H50NNTX3BBCFG":3,"skill-same:01M6BY2W6MYA5H50NNTX3BBCFG":102},{"skillNo":4,"name":5,"nameZh":6,"description":7,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":16,"categories":17,"keywords":23,"publishedAt":39,"linkOnly":40,"markdownContent":41,"markdownBody":42,"frontMatter":43,"refName":45,"commitSha":46,"skillDirectory":47,"forksCount":48,"repositoryHtmlUrl":49,"repositoryHomepageUrl":50,"ownerHtmlUrl":51,"ownerBlogUrl":12,"ownerDescription":52,"updatedAt":53,"install":54,"resources":66},"01M6BY2W6MYA5H50NNTX3BBCFG","nodetool-workflow-builder",null,"Create or edit NodeTool workflow graphs, connections, and properties using UI, headless tools, or requested workflow files.","nodetool-ai\u002Fnodetool","nodetool-ai","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F164873569?v=4","01M6BXPFDW36K9K1MR8ZH63HG2","nodetool.ai","01M6BXPFJ893XM219GQN60B0WK",635,"AGPL-3.0","https:\u002F\u002Fgithub.com\u002Fnodetool-ai\u002Fnodetool\u002Fblob\u002Fmain\u002Fpackages\u002Fsystem-skills\u002Fnodetool-workflow-builder\u002FSKILL.md",[18],{"categoryNo":19,"categoryName":20,"taxonomyType":21,"taxonomyTypeDesc":22},"01JBSEEDCAT000000000000023","Skill与智能体开发","SCENARIO","使用场景",[24,25,26,27,28,29,30,31,32,33,34,35,36,37,38],"workflow","nodetool workflow builder","nodetool","workflow builder","builder","create","ui","Tools","edit","properties","files","connections","graphs","true","headless","2026-11-10 10:12:30",false,"---\nname: nodetool-workflow-builder\ndescription: \"Create or edit NodeTool workflow graphs, connections, and properties using UI, headless tools, or requested workflow files.\"\nfeatured: true\n---\n\n# Build NodeTool workflows\n\nCreate or edit a workflow that produces the requested outputs from the specified\ninputs. Preserve existing nodes and connections outside the requested change.\n\n## A graph is one surface among several\n\nNodeTool holds several document kinds, and most of them are not graphs. Build a\ngraph when the user asked for one, when the job has to re-run on new inputs, or\nwhen the result must be callable over the API. Otherwise route the request.\n\n| The deliverable | Skill |\n|---|---|\n| A finished video, still set or campaign | `storyboard-core`, which picks the job skill |\n| A screen someone clicks | `nodetool-app-builder` |\n| A repair on footage already cut: cutout, upscale, lip sync, generated sound | `nodetool-video-post` |\n| A cut, titles, transitions or animation on a timeline | the `motion-graphics` system skill, via `load_skill` |\n| Logic in JavaScript rather than nodes | `nodetool-js-scripting` |\n| A layered image, mask or overlay | `nodetool-sketch` |\n| A 3D model or scene | `nodetool-3d-scene` |\n| A playable 2D or 3D game | `native-game` |\n| A math or explainer animation in Python | `manimgx` |\n| Document indexing, vector search, RAG | `nodetool-rag-indexing` |\n| Navigating, reading or filling in web pages | `nodetool-browser-agent` |\n| A reusable video template for batches | `video-workflow` |\n| A new node type | `nodetool-custom-node-developer` |\n\nA workflow can still be the right answer alongside one of these: a mini app runs\nworkflows, and a timeline clip can be produced by one.\n\nUse the authoring surface the user requested: the graph tools for a saved or\nopen workflow, and JSON or the TypeScript DSL when a workflow file is\nrequested. Do not claim a missing tool is callable. Use the live schemas and\nregistry as the source of identifiers.\n\n## Build and verify\n\n1. Inspect the existing graph for edits (`ui_get_graph` or `get_workflow`).\n   Identify required inputs, outputs, and behavior from the request.\n2. Find node types. On the canvas, `ui_search_nodes` with\n   `include_properties: true` and `include_outputs: true` returns property and\n   output names. Headlessly, `search_nodes {query: [...]}` finds the type and\n   `get_node_info {node_type}` returns its inputs, outputs, types and defaults.\n   Pick models with `find_model` (headless) or `ui_search_models` (canvas). Do\n   not invent node types, properties, handles, or model identifiers.\n3. Add or update nodes, set required properties, and connect compatible handles.\n   Keep existing layout and identifiers when editing.\n4. Validate with `ui_get_graph` (its `validation` field) or `validate_workflow`.\n   Fix errors and required-input warnings. Read back the saved graph before\n   claiming it was created or changed. Report any remaining limitation.\n5. Execute only when the request includes a run. A run of generation nodes\n   spends provider money, so confirm scope with the user first. Report the\n   workflow id, inputs, validation result, and actual run outcome.\n\nIf a node is unavailable, broaden the search and inspect supported alternatives.\nProceed with an equivalent composition when it preserves the requested behavior.\nOtherwise explain the missing capability and ask only for the decision needed to\nchoose an alternative. Continue independent work on the rest of the graph.\n\n## Graph tools\n\nThese edit one workflow. With the workflow open in the editor they act on the\nlive canvas. With no editor open, the eight graph tools (`ui_get_graph`,\n`ui_add_node`, `ui_connect_nodes`, `ui_update_node_data`, `ui_delete_node`,\n`ui_delete_edge`, `ui_move_node`, `ui_set_node_title`) read and write the saved\nworkflow named by `workflow_id`. Each mutation also accepts\n`based_on_revision` (from `ui_get_graph`), which refuses the edit if the graph\nchanged.\n\n| Tool | Purpose | Key params |\n|------|---------|-----------|\n| `ui_search_nodes` | Find node types | `query`, `input_type`, `output_type`, `include_properties`, `include_outputs`, `limit` |\n| `ui_search_models` | Models for a generic AI node | `kind` (`text_to_image`, `image_to_image`, `text_to_video`, `image_to_video`, `reference_to_video`, `text_to_speech`, `text_to_music`, `audio_to_audio`, `speech_to_text`, `text_generation`, `embedding`), `limit` |\n| `ui_add_node` | Add one node | `id`, `type`, `position {x, y}`, `properties` |\n| `ui_connect_nodes` | Connect two nodes | `source_node_id`, `source_handle`, `target_node_id`, `target_handle` |\n| `ui_update_node_data` | Merge into a node's `data` | `node_id`, `data` (e.g. `{properties: {…}}`, `{dynamic_outputs: {…}}`, `{title}`) |\n| `ui_get_graph` | Read graph, revision and validation | `workflow_id` |\n| `ui_delete_node` \u002F `ui_delete_edge` | Remove | `node_id` \u002F `edge_id` |\n| `ui_move_node` | Reposition | `node_id`, `position {x, y}` |\n| `ui_set_node_title` | Rename a node | `node_id`, `title` |\n| `ui_open_workflow` | Open a workflow tab | `workflow_id` |\n| `ui_run_workflow` | Execute | `workflow_id`, `params` |\n\n`ui_add_node` takes `properties` only. Anything else on the node's `data`, such\nas `dynamic_outputs`, is set afterwards with `ui_update_node_data`.\n\n## Workflow tools\n\nThese need no editor.\n\n| Tool | Purpose |\n|---|---|\n| `list_workflows`, `get_workflow` | Find and read a workflow |\n| `search_nodes`, `get_node_info`, `list_nodes` | Node discovery. `search_nodes` hides provider nodes (`openai.*`, `kie.*`, …) unless `include_provider_nodes: true`, which is for a provider the user named |\n| `find_model` | A `{provider, model}` ref for a generic AI node's `model` property |\n| `create_workflow {name, graph}` | Save a new graph. Nodes are `{id, type, properties}`. Edges are `{source, sourceHandle, target, targetHandle}` |\n| `validate_workflow {workflow_id \\| graph}` | Static check: unknown types, missing properties, unselected models, bad edges |\n| `run_workflow`, `start_background_job`, `debug_workflow` | Execute, execute without waiting, or execute with a full debug report |\n| `run_node {node_type, inputs}` | Run one node to probe what it does (chat only) |\n\nFor a brief rather than a graph, the planner builds one: `plan_workflow\n{workflow_id}` stores a plan whose steps map to registry node types,\n`update_workflow_plan_step` fixes a step the registry lacks, and\n`build_workflow_from_plan {workflow_id}` places and validates the nodes. On the\ncanvas the same flow is `ui_workflow_plan`, `ui_workflow_update_plan_step` and\n`ui_workflow_build_from_plan`.\n\n## Node catalog\n\nCatalog entries are search leads. Confirm each type with `get_node_info` or\n`ui_search_nodes` before using it. From the CLI, inspect installed types with\n`nodetool node list` and `nodetool node info \u003Ctype>`.\n\n`nodetool.output.Output` is supplied by the audio pack. If it is needed but\nnot installed, run `nodetool packs add audio --yes`. CLI workflow runs also\nreturn terminal transform outputs: a core-only deterministic text graph can\nend at `nodetool.text.Template` without a dedicated sink. For example,\n`StringInput` → `Template` returns the template's values under `outputs.Template`.\n\n| Namespace | Common nodes |\n|---|---|\n| `nodetool.input` | `StringInput`, `IntegerInput`, `FloatInput`, `BooleanInput`, `ImageInput`, `AudioInput`, `VideoInput`, `DocumentInput`, `MessageInput`, `LanguageModelInput` |\n| `nodetool.output` | `Output`. `nodetool.workflows.base_node.Preview` shows an intermediate value |\n| `nodetool.constant` | `String`, `Integer`, `Float`, `Bool`, `List`, `Dict`, `Image`, `Audio`, `Video`, `JSON` |\n| `nodetool.agents` | `Agent`, `Summarizer`, `Extractor`, `Classifier`, `Decision`, `EnhancePrompt`, `CreateThread` |\n| `nodetool.generators` | `ListGenerator`, `DataGenerator`, `StructuredOutputGenerator`, `ChartGenerator`, `SVGGenerator` |\n| `nodetool.text` | `Template`, `Concat`, `Prompt`, `FilterString`, `FilterRegexString`, `Embedding`, `AutomaticSpeechRecognition`, `SaveText` |\n| `nodetool.code` | `Code`: a JavaScript body. See `nodetool-js-scripting` |\n| `nodetool.control` | `If`, `Switch`, `ForEach`, `Collect`, `Loop`, `Chunk`, `Take`, `Zip`, `RepeatCount`, `RepeatValue`, `TryCatch` |\n| `nodetool.data` | `ForEachRow`, `LoadCSVAssets` |\n| `nodetool.document` | `LoadDocumentFile`, `SaveDocumentFile`, `ListDocuments` |\n| `nodetool.image` | `TextToImage`, `ImageToImage`, `LoadImageFile`, `SaveImage`, `Resize`, `Crop`, `RotateAndFlip`, `Compositor`, `Upscale`, `RemoveBackground` |\n| `nodetool.audio` | `TextToSpeech`, `TextToMusic`, `LoadAudioFile`, `SaveAudio`, `AudioMixer`, `Concat`, `Trim`, `Normalize` |\n| `nodetool.video` | `TextToVideo`, `ImageToVideo`, `ReferenceToVideo`, `LoadVideoFile`, `SaveVideo`, `Concat`, `Trim`, `ExtractFrame`, `ExtractAudio`, `AddAudio`, `GetVideoInfo` |\n| `vector` | `Collection`, `IndexTextChunk`, `QueryText`, `HybridSearch`. See `nodetool-rag-indexing` |\n| `lib.*` | `lib.pdf` (text, tables, page images), `lib.image.*` (filters, color, draw), `lib.audio.*` (effects), `lib.svg`, `lib.charts.ChartRenderer`, `lib.browser.Screenshot`, `lib.sqlite.GetDatabasePath` |\n\n`lib.http`, `lib.os`, `lib.datetime`, `lib.json`, `lib.math`, `lib.uuid`,\n`nodetool.boolean`, `nodetool.dictionary`, `nodetool.list`, `nodetool.numbers`,\nthe old `nodetool.text` helpers such as `Slugify` and `Chunk`, and all `skills.*`\nnodes have been removed. Use the Code node (`nodetool.code.Code`) and its\nsnippet library instead. Its body is plain JavaScript with `fetch`,\n`workspace.*` and the other globals `nodetool-js-scripting` lists. Libraries\nsuch as `@nodetool-ai\u002Fsandbox-dates` are imported sandbox packages, not globals.\n\n### Generation nodes\n\nUse the generic nodes with a `model` property: `nodetool.image.TextToImage`,\n`ImageToImage`, `nodetool.video.TextToVideo`, `ImageToVideo`,\n`ReferenceToVideo`, `nodetool.audio.TextToSpeech`, `TextToMusic`. Set `model`\nfrom `find_model` or `ui_search_models`. A provider's own node (`kie.video.*`,\n`openai.*`, `gemini.*`, …) exposes that model's extra settings. Search for one\nonly when the user named the provider or needs a setting the generic node does\nnot carry, with `include_provider_nodes: true` on `search_nodes`.\n\n### Data types\n\n- **Primitives**: `str`, `int`, `float`, `bool`, `list`, `dict`, `any`\n- **Assets**: `{type: \"image|audio|video|document\", uri: \"...\"}`\n- **Models**: `language_model`, `image_model`, `video_model`, `embedding_model`, `tts_model`\n\nEdges enforce type compatibility. Use `any` for flexible connections. Search\nwith `input_type` \u002F `output_type` (`\"str\"`, `\"image\"`, `\"list\"`, …) to find a\nnode that fits a handle.\n\n## Workflow patterns\n\n- **Pipeline.** Input → transforms → Output. For example `ImageInput` →\n  `lib.image.enhance.AutoContrast` → `SaveImage` → `Output`.\n- **Agent generation.** Input → `Agent` or `Summarizer` → post-process →\n  Output. `ListGenerator` streams one item per line for fan-out.\n- **Previews.** Put `Preview` nodes after intermediate stages of a multi-stage\n  generation so progress is visible.\n- **RAG.** Index: Code node (`workspace.list`) → `LoadDocumentFile` → Code node\n  (\"Chunk Text\" snippet) → `vector.IndexTextChunk`. Query: `StringInput` →\n  `vector.HybridSearch` → `Template` → `Agent` → `Output`. See\n  `nodetool-rag-indexing`.\n- **Realtime.** `RealtimeAudioInput` → `openai.agents.RealtimeAgent` or\n  `openai.agents.LiveAgent` → `Preview`. `LiveAgent` runs full-duplex speech;\n  feed it 24 kHz PCM16 microphone chunks.\n- **Multi-modal.** Chain modalities: `AutomaticSpeechRecognition` → `Agent` →\n  `TextToImage`, or `ExtractAudio` → `AutomaticSpeechRecognition` →\n  `Summarizer`.\n- **Generated media.** `StringInput` (prompt) → `TextToVideo` → `Output`, or\n  `ImageInput` + `StringInput` → `ImageToVideo` → `Output`. For lip sync,\n  upscaling or avatar work, search `\"lip sync\"`, `\"upscale\"` or `\"avatar\"`.\n\n### Agent outputs and tools\n\n- `dynamic_outputs` on an `Agent` (`{\"title\": {\"type\": \"str\"}}`) declares\n  structured result fields. The model fills them through a submit tool, and each\n  becomes an output handle.\n- To let an Agent call another node as a tool, connect it with a control edge\n  (`edge_type: \"control\"`, both handles `__control__`). The Agent's regular\n  outputs (`text`, `chunk`) still route to `Preview` or `Output`.\n\n### Streaming and control flow\n\n- Everything is a stream. A single value is a one-item stream.\n- `nodetool.control.Collect` gathers a stream into a list.\n  `nodetool.control.ForEach` emits each item of a list.\n- `nodetool.control.Loop` repeats part of the graph: wire `initial` into the\n  body, the body's result back into `next`, and a bool into `condition` (loop\n  again while true). `max_iterations` bounds it and `done` carries the final\n  value. It is the only way to close a cycle. The DSL cannot express the back\n  edge, so build loops with the graph tools or workflow JSON.\n- `nodetool.agents.Decision` lets a model make the call: `prompt` is a yes\u002Fno\n  question, `value` is what it judges (images are shown to the model),\n  `decision` is the bool and `if_true`\u002F`if_false` route `value` like `If`. Wire\n  `decision` into `Loop.condition` to loop until a model is satisfied, and phrase\n  the question so yes means \"loop again\".\n\n## Debugging and validation\n\n`ui_get_graph` returns a `validation` field:\n\n- `errors`: blocking issues, such as invalid node types or a Code body that does\n  not parse.\n- `warnings`: disconnected required inputs and empty required properties.\n- `suggestions`: orphaned nodes.\n\n`validate_workflow` is the headless equivalent and also catches unselected\nmodels and model ids a provider does not offer. Fix blocking errors and the\nwarnings that prevent the requested behavior. Explain any remaining warning\nwithout expanding the task.\n\n- **\"Required property 'X' is not set\"**: set it with `ui_update_node_data` or in\n  the graph's `properties`. Common ones are `model` and `prompt`.\n- **\"Required input 'X' not connected\"**: add an edge or set a value.\n- **Wrong handle name**: re-read the type with `get_node_info` or\n  `ui_search_nodes {include_outputs: true}`.\n- **Type mismatch**: compare the source output type with the target input type.\n- **Node not found**: broaden the query or search by namespace prefix\n  (`\"nodetool.text\"`).\n\nDo not retry a failing call unchanged. Agent and generator nodes need `model`,\ngeneration nodes need `prompt`, and `ForEach` needs a list input. Set them or ask\nthe user which value to use.\n\n## Running workflows from a shell\n\nThis section is for a person or agent with a shell. The in-product agent uses\n`run_workflow` instead.\n\n```bash\nnodetool workflows run .\u002Fworkflow.json --params '{\"input\": \"hello\"}' --json\nnodetool workflows run .\u002Fworkflow.ts --json    # a DSL file, built then run\nnodetool workflows run \u003Cworkflow_id>\nnodetool workflows list\nnodetool workflows get \u003Cid>\nnodetool jobs list\nnodetool serve --port 7777\n\n# From a NodeTool checkout\nnpm run workflow -- .\u002Fworkflow.json --input text='hello' --show-messages\n```\n\n## TypeScript DSL\n\nWrite workflows as TypeScript with `@nodetool-ai\u002Fdsl`. Namespaces mirror node\nnamespaces: `nodetool.text` is `text`, `nodetool.control` is `control`,\n`lib.pdf` is `libPdf`, `vector` is `vector`.\n\n```typescript\nimport { workflow, input, text, agents, output } from \"@nodetool-ai\u002Fdsl\";\n\nconst topic = input.stringInput({ name: \"topic\", value: \"tides\" });\nconst brief = text.template({\n  string: \"Explain {{ topic }} to a ten-year-old.\",\n  topic: topic.output()\n});\nconst answer = agents.summarizer({ text: brief.output() });\nconst result = output.output({ name: \"answer\", value: answer.output(\"text\") });\n\nconst wf = workflow(result);\nconsole.log(JSON.stringify(wf));\n```\n\n- **`node.output()`** is a function. It returns the default output handle and\n  only works on a node with one. Name the slot on a multi-output node:\n  `answer.output(\"text\")`, `branch.output(\"if_true\")`.\n- **Connectable**: every input accepts a literal value or an output handle.\n- **`workflow(...terminals)`** traces from the terminal nodes and returns\n  serializable `{nodes, edges}`. A node shared by two branches appears once.\n- **`run(wf, opts?)`** executes in-process. **`runGraph(...terminals)`** is\n  `run(workflow(...terminals))`.\n- Dynamic inputs are extra keys, as `topic` on `text.template` above.\n- There is no `libMath`, `libHttp` or `list` namespace. Use `code.code` for\n  arithmetic, JSON, list and HTTP logic.\n","# Build NodeTool workflows\n\nCreate or edit a workflow that produces the requested outputs from the specified\ninputs. Preserve existing nodes and connections outside the requested change.\n\n## A graph is one surface among several\n\nNodeTool holds several document kinds, and most of them are not graphs. Build a\ngraph when the user asked for one, when the job has to re-run on new inputs, or\nwhen the result must be callable over the API. Otherwise route the request.\n\n| The deliverable | Skill |\n|---|---|\n| A finished video, still set or campaign | `storyboard-core`, which picks the job skill |\n| A screen someone clicks | `nodetool-app-builder` |\n| A repair on footage already cut: cutout, upscale, lip sync, generated sound | `nodetool-video-post` |\n| A cut, titles, transitions or animation on a timeline | the `motion-graphics` system skill, via `load_skill` |\n| Logic in JavaScript rather than nodes | `nodetool-js-scripting` |\n| A layered image, mask or overlay | `nodetool-sketch` |\n| A 3D model or scene | `nodetool-3d-scene` |\n| A playable 2D or 3D game | `native-game` |\n| A math or explainer animation in Python | `manimgx` |\n| Document indexing, vector search, RAG | `nodetool-rag-indexing` |\n| Navigating, reading or filling in web pages | `nodetool-browser-agent` |\n| A reusable video template for batches | `video-workflow` |\n| A new node type | `nodetool-custom-node-developer` |\n\nA workflow can still be the right answer alongside one of these: a mini app runs\nworkflows, and a timeline clip can be produced by one.\n\nUse the authoring surface the user requested: the graph tools for a saved or\nopen workflow, and JSON or the TypeScript DSL when a workflow file is\nrequested. Do not claim a missing tool is callable. Use the live schemas and\nregistry as the source of identifiers.\n\n## Build and verify\n\n1. Inspect the existing graph for edits (`ui_get_graph` or `get_workflow`).\n   Identify required inputs, outputs, and behavior from the request.\n2. Find node types. On the canvas, `ui_search_nodes` with\n   `include_properties: true` and `include_outputs: true` returns property and\n   output names. Headlessly, `search_nodes {query: [...]}` finds the type and\n   `get_node_info {node_type}` returns its inputs, outputs, types and defaults.\n   Pick models with `find_model` (headless) or `ui_search_models` (canvas). Do\n   not invent node types, properties, handles, or model identifiers.\n3. Add or update nodes, set required properties, and connect compatible handles.\n   Keep existing layout and identifiers when editing.\n4. Validate with `ui_get_graph` (its `validation` field) or `validate_workflow`.\n   Fix errors and required-input warnings. Read back the saved graph before\n   claiming it was created or changed. Report any remaining limitation.\n5. Execute only when the request includes a run. A run of generation nodes\n   spends provider money, so confirm scope with the user first. Report the\n   workflow id, inputs, validation result, and actual run outcome.\n\nIf a node is unavailable, broaden the search and inspect supported alternatives.\nProceed with an equivalent composition when it preserves the requested behavior.\nOtherwise explain the missing capability and ask only for the decision needed to\nchoose an alternative. Continue independent work on the rest of the graph.\n\n## Graph tools\n\nThese edit one workflow. With the workflow open in the editor they act on the\nlive canvas. With no editor open, the eight graph tools (`ui_get_graph`,\n`ui_add_node`, `ui_connect_nodes`, `ui_update_node_data`, `ui_delete_node`,\n`ui_delete_edge`, `ui_move_node`, `ui_set_node_title`) read and write the saved\nworkflow named by `workflow_id`. Each mutation also accepts\n`based_on_revision` (from `ui_get_graph`), which refuses the edit if the graph\nchanged.\n\n| Tool | Purpose | Key params |\n|------|---------|-----------|\n| `ui_search_nodes` | Find node types | `query`, `input_type`, `output_type`, `include_properties`, `include_outputs`, `limit` |\n| `ui_search_models` | Models for a generic AI node | `kind` (`text_to_image`, `image_to_image`, `text_to_video`, `image_to_video`, `reference_to_video`, `text_to_speech`, `text_to_music`, `audio_to_audio`, `speech_to_text`, `text_generation`, `embedding`), `limit` |\n| `ui_add_node` | Add one node | `id`, `type`, `position {x, y}`, `properties` |\n| `ui_connect_nodes` | Connect two nodes | `source_node_id`, `source_handle`, `target_node_id`, `target_handle` |\n| `ui_update_node_data` | Merge into a node's `data` | `node_id`, `data` (e.g. `{properties: {…}}`, `{dynamic_outputs: {…}}`, `{title}`) |\n| `ui_get_graph` | Read graph, revision and validation | `workflow_id` |\n| `ui_delete_node` \u002F `ui_delete_edge` | Remove | `node_id` \u002F `edge_id` |\n| `ui_move_node` | Reposition | `node_id`, `position {x, y}` |\n| `ui_set_node_title` | Rename a node | `node_id`, `title` |\n| `ui_open_workflow` | Open a workflow tab | `workflow_id` |\n| `ui_run_workflow` | Execute | `workflow_id`, `params` |\n\n`ui_add_node` takes `properties` only. Anything else on the node's `data`, such\nas `dynamic_outputs`, is set afterwards with `ui_update_node_data`.\n\n## Workflow tools\n\nThese need no editor.\n\n| Tool | Purpose |\n|---|---|\n| `list_workflows`, `get_workflow` | Find and read a workflow |\n| `search_nodes`, `get_node_info`, `list_nodes` | Node discovery. `search_nodes` hides provider nodes (`openai.*`, `kie.*`, …) unless `include_provider_nodes: true`, which is for a provider the user named |\n| `find_model` | A `{provider, model}` ref for a generic AI node's `model` property |\n| `create_workflow {name, graph}` | Save a new graph. Nodes are `{id, type, properties}`. Edges are `{source, sourceHandle, target, targetHandle}` |\n| `validate_workflow {workflow_id \\| graph}` | Static check: unknown types, missing properties, unselected models, bad edges |\n| `run_workflow`, `start_background_job`, `debug_workflow` | Execute, execute without waiting, or execute with a full debug report |\n| `run_node {node_type, inputs}` | Run one node to probe what it does (chat only) |\n\nFor a brief rather than a graph, the planner builds one: `plan_workflow\n{workflow_id}` stores a plan whose steps map to registry node types,\n`update_workflow_plan_step` fixes a step the registry lacks, and\n`build_workflow_from_plan {workflow_id}` places and validates the nodes. On the\ncanvas the same flow is `ui_workflow_plan`, `ui_workflow_update_plan_step` and\n`ui_workflow_build_from_plan`.\n\n## Node catalog\n\nCatalog entries are search leads. Confirm each type with `get_node_info` or\n`ui_search_nodes` before using it. From the CLI, inspect installed types with\n`nodetool node list` and `nodetool node info \u003Ctype>`.\n\n`nodetool.output.Output` is supplied by the audio pack. If it is needed but\nnot installed, run `nodetool packs add audio --yes`. CLI workflow runs also\nreturn terminal transform outputs: a core-only deterministic text graph can\nend at `nodetool.text.Template` without a dedicated sink. For example,\n`StringInput` → `Template` returns the template's values under `outputs.Template`.\n\n| Namespace | Common nodes |\n|---|---|\n| `nodetool.input` | `StringInput`, `IntegerInput`, `FloatInput`, `BooleanInput`, `ImageInput`, `AudioInput`, `VideoInput`, `DocumentInput`, `MessageInput`, `LanguageModelInput` |\n| `nodetool.output` | `Output`. `nodetool.workflows.base_node.Preview` shows an intermediate value |\n| `nodetool.constant` | `String`, `Integer`, `Float`, `Bool`, `List`, `Dict`, `Image`, `Audio`, `Video`, `JSON` |\n| `nodetool.agents` | `Agent`, `Summarizer`, `Extractor`, `Classifier`, `Decision`, `EnhancePrompt`, `CreateThread` |\n| `nodetool.generators` | `ListGenerator`, `DataGenerator`, `StructuredOutputGenerator`, `ChartGenerator`, `SVGGenerator` |\n| `nodetool.text` | `Template`, `Concat`, `Prompt`, `FilterString`, `FilterRegexString`, `Embedding`, `AutomaticSpeechRecognition`, `SaveText` |\n| `nodetool.code` | `Code`: a JavaScript body. See `nodetool-js-scripting` |\n| `nodetool.control` | `If`, `Switch`, `ForEach`, `Collect`, `Loop`, `Chunk`, `Take`, `Zip`, `RepeatCount`, `RepeatValue`, `TryCatch` |\n| `nodetool.data` | `ForEachRow`, `LoadCSVAssets` |\n| `nodetool.document` | `LoadDocumentFile`, `SaveDocumentFile`, `ListDocuments` |\n| `nodetool.image` | `TextToImage`, `ImageToImage`, `LoadImageFile`, `SaveImage`, `Resize`, `Crop`, `RotateAndFlip`, `Compositor`, `Upscale`, `RemoveBackground` |\n| `nodetool.audio` | `TextToSpeech`, `TextToMusic`, `LoadAudioFile`, `SaveAudio`, `AudioMixer`, `Concat`, `Trim`, `Normalize` |\n| `nodetool.video` | `TextToVideo`, `ImageToVideo`, `ReferenceToVideo`, `LoadVideoFile`, `SaveVideo`, `Concat`, `Trim`, `ExtractFrame`, `ExtractAudio`, `AddAudio`, `GetVideoInfo` |\n| `vector` | `Collection`, `IndexTextChunk`, `QueryText`, `HybridSearch`. See `nodetool-rag-indexing` |\n| `lib.*` | `lib.pdf` (text, tables, page images), `lib.image.*` (filters, color, draw), `lib.audio.*` (effects), `lib.svg`, `lib.charts.ChartRenderer`, `lib.browser.Screenshot`, `lib.sqlite.GetDatabasePath` |\n\n`lib.http`, `lib.os`, `lib.datetime`, `lib.json`, `lib.math`, `lib.uuid`,\n`nodetool.boolean`, `nodetool.dictionary`, `nodetool.list`, `nodetool.numbers`,\nthe old `nodetool.text` helpers such as `Slugify` and `Chunk`, and all `skills.*`\nnodes have been removed. Use the Code node (`nodetool.code.Code`) and its\nsnippet library instead. Its body is plain JavaScript with `fetch`,\n`workspace.*` and the other globals `nodetool-js-scripting` lists. Libraries\nsuch as `@nodetool-ai\u002Fsandbox-dates` are imported sandbox packages, not globals.\n\n### Generation nodes\n\nUse the generic nodes with a `model` property: `nodetool.image.TextToImage`,\n`ImageToImage`, `nodetool.video.TextToVideo`, `ImageToVideo`,\n`ReferenceToVideo`, `nodetool.audio.TextToSpeech`, `TextToMusic`. Set `model`\nfrom `find_model` or `ui_search_models`. A provider's own node (`kie.video.*`,\n`openai.*`, `gemini.*`, …) exposes that model's extra settings. Search for one\nonly when the user named the provider or needs a setting the generic node does\nnot carry, with `include_provider_nodes: true` on `search_nodes`.\n\n### Data types\n\n- **Primitives**: `str`, `int`, `float`, `bool`, `list`, `dict`, `any`\n- **Assets**: `{type: \"image|audio|video|document\", uri: \"...\"}`\n- **Models**: `language_model`, `image_model`, `video_model`, `embedding_model`, `tts_model`\n\nEdges enforce type compatibility. Use `any` for flexible connections. Search\nwith `input_type` \u002F `output_type` (`\"str\"`, `\"image\"`, `\"list\"`, …) to find a\nnode that fits a handle.\n\n## Workflow patterns\n\n- **Pipeline.** Input → transforms → Output. For example `ImageInput` →\n  `lib.image.enhance.AutoContrast` → `SaveImage` → `Output`.\n- **Agent generation.** Input → `Agent` or `Summarizer` → post-process →\n  Output. `ListGenerator` streams one item per line for fan-out.\n- **Previews.** Put `Preview` nodes after intermediate stages of a multi-stage\n  generation so progress is visible.\n- **RAG.** Index: Code node (`workspace.list`) → `LoadDocumentFile` → Code node\n  (\"Chunk Text\" snippet) → `vector.IndexTextChunk`. Query: `StringInput` →\n  `vector.HybridSearch` → `Template` → `Agent` → `Output`. See\n  `nodetool-rag-indexing`.\n- **Realtime.** `RealtimeAudioInput` → `openai.agents.RealtimeAgent` or\n  `openai.agents.LiveAgent` → `Preview`. `LiveAgent` runs full-duplex speech;\n  feed it 24 kHz PCM16 microphone chunks.\n- **Multi-modal.** Chain modalities: `AutomaticSpeechRecognition` → `Agent` →\n  `TextToImage`, or `ExtractAudio` → `AutomaticSpeechRecognition` →\n  `Summarizer`.\n- **Generated media.** `StringInput` (prompt) → `TextToVideo` → `Output`, or\n  `ImageInput` + `StringInput` → `ImageToVideo` → `Output`. For lip sync,\n  upscaling or avatar work, search `\"lip sync\"`, `\"upscale\"` or `\"avatar\"`.\n\n### Agent outputs and tools\n\n- `dynamic_outputs` on an `Agent` (`{\"title\": {\"type\": \"str\"}}`) declares\n  structured result fields. The model fills them through a submit tool, and each\n  becomes an output handle.\n- To let an Agent call another node as a tool, connect it with a control edge\n  (`edge_type: \"control\"`, both handles `__control__`). The Agent's regular\n  outputs (`text`, `chunk`) still route to `Preview` or `Output`.\n\n### Streaming and control flow\n\n- Everything is a stream. A single value is a one-item stream.\n- `nodetool.control.Collect` gathers a stream into a list.\n  `nodetool.control.ForEach` emits each item of a list.\n- `nodetool.control.Loop` repeats part of the graph: wire `initial` into the\n  body, the body's result back into `next`, and a bool into `condition` (loop\n  again while true). `max_iterations` bounds it and `done` carries the final\n  value. It is the only way to close a cycle. The DSL cannot express the back\n  edge, so build loops with the graph tools or workflow JSON.\n- `nodetool.agents.Decision` lets a model make the call: `prompt` is a yes\u002Fno\n  question, `value` is what it judges (images are shown to the model),\n  `decision` is the bool and `if_true`\u002F`if_false` route `value` like `If`. Wire\n  `decision` into `Loop.condition` to loop until a model is satisfied, and phrase\n  the question so yes means \"loop again\".\n\n## Debugging and validation\n\n`ui_get_graph` returns a `validation` field:\n\n- `errors`: blocking issues, such as invalid node types or a Code body that does\n  not parse.\n- `warnings`: disconnected required inputs and empty required properties.\n- `suggestions`: orphaned nodes.\n\n`validate_workflow` is the headless equivalent and also catches unselected\nmodels and model ids a provider does not offer. Fix blocking errors and the\nwarnings that prevent the requested behavior. Explain any remaining warning\nwithout expanding the task.\n\n- **\"Required property 'X' is not set\"**: set it with `ui_update_node_data` or in\n  the graph's `properties`. Common ones are `model` and `prompt`.\n- **\"Required input 'X' not connected\"**: add an edge or set a value.\n- **Wrong handle name**: re-read the type with `get_node_info` or\n  `ui_search_nodes {include_outputs: true}`.\n- **Type mismatch**: compare the source output type with the target input type.\n- **Node not found**: broaden the query or search by namespace prefix\n  (`\"nodetool.text\"`).\n\nDo not retry a failing call unchanged. Agent and generator nodes need `model`,\ngeneration nodes need `prompt`, and `ForEach` needs a list input. Set them or ask\nthe user which value to use.\n\n## Running workflows from a shell\n\nThis section is for a person or agent with a shell. The in-product agent uses\n`run_workflow` instead.\n\n```bash\nnodetool workflows run .\u002Fworkflow.json --params '{\"input\": \"hello\"}' --json\nnodetool workflows run .\u002Fworkflow.ts --json    # a DSL file, built then run\nnodetool workflows run \u003Cworkflow_id>\nnodetool workflows list\nnodetool workflows get \u003Cid>\nnodetool jobs list\nnodetool serve --port 7777\n\n# From a NodeTool checkout\nnpm run workflow -- .\u002Fworkflow.json --input text='hello' --show-messages\n```\n\n## TypeScript DSL\n\nWrite workflows as TypeScript with `@nodetool-ai\u002Fdsl`. Namespaces mirror node\nnamespaces: `nodetool.text` is `text`, `nodetool.control` is `control`,\n`lib.pdf` is `libPdf`, `vector` is `vector`.\n\n```typescript\nimport { workflow, input, text, agents, output } from \"@nodetool-ai\u002Fdsl\";\n\nconst topic = input.stringInput({ name: \"topic\", value: \"tides\" });\nconst brief = text.template({\n  string: \"Explain {{ topic }} to a ten-year-old.\",\n  topic: topic.output()\n});\nconst answer = agents.summarizer({ text: brief.output() });\nconst result = output.output({ name: \"answer\", value: answer.output(\"text\") });\n\nconst wf = workflow(result);\nconsole.log(JSON.stringify(wf));\n```\n\n- **`node.output()`** is a function. It returns the default output handle and\n  only works on a node with one. Name the slot on a multi-output node:\n  `answer.output(\"text\")`, `branch.output(\"if_true\")`.\n- **Connectable**: every input accepts a literal value or an output handle.\n- **`workflow(...terminals)`** traces from the terminal nodes and returns\n  serializable `{nodes, edges}`. A node shared by two branches appears once.\n- **`run(wf, opts?)`** executes in-process. **`runGraph(...terminals)`** is\n  `run(workflow(...terminals))`.\n- Dynamic inputs are extra keys, as `topic` on `text.template` above.\n- There is no `libMath`, `libHttp` or `list` namespace. Use `code.code` for\n  arithmetic, JSON, list and HTTP logic.\n",{"name":5,"featured":44,"description":7},true,"main","c7226baa82daca42a19a5855c116544f97aa8bac","packages\u002Fsystem-skills\u002Fnodetool-workflow-builder",91,"https:\u002F\u002Fgithub.com\u002Fnodetool-ai\u002Fnodetool","https:\u002F\u002Fnodetool.ai\u002F","https:\u002F\u002Fgithub.com\u002Fnodetool-ai","No-code AI development","2026-11-15 03:27:49",{"source":8,"skillName":5,"commands":55,"prompt":65,"zipAvailable":44},[56,59,62],{"tool":57,"command":58},"npx","npx skills add nodetool-ai\u002Fnodetool --skill nodetool-workflow-builder",{"tool":60,"command":61},"bunx","bunx skills add nodetool-ai\u002Fnodetool --skill nodetool-workflow-builder",{"tool":63,"command":64},"pnpm","pnpm dlx skills add nodetool-ai\u002Fnodetool --skill nodetool-workflow-builder","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002Fnodetool-ai\u002Fnodetool\u002Fblob\u002Fmain\u002Fpackages\u002Fsystem-skills\u002Fnodetool-workflow-builder\u002FSKILL.md",[67,73,77,81,85,89,94,98],{"path":68,"resourceType":69,"resourceTypeDesc":70,"sizeBytes":71,"sourceHtmlUrl":72},"agents\u002Fopenai.yaml","REFERENCE","参考文档",113,"https:\u002F\u002Fgithub.com\u002Fboisenoise\u002Fskills-collections\u002Fblob\u002Fmain\u002Fskills\u002Fantigravity-better-ui\u002Fagents\u002Fopenai.yaml",{"path":74,"resourceType":69,"resourceTypeDesc":70,"sizeBytes":75,"sourceHtmlUrl":76},"animations.md",4923,"https:\u002F\u002Fgithub.com\u002Fboisenoise\u002Fskills-collections\u002Fblob\u002Fmain\u002Fskills\u002Fantigravity-better-ui\u002Fanimations.md",{"path":78,"resourceType":69,"resourceTypeDesc":70,"sizeBytes":79,"sourceHtmlUrl":80},"enter-exit.md",3596,"https:\u002F\u002Fgithub.com\u002Fboisenoise\u002Fskills-collections\u002Fblob\u002Fmain\u002Fskills\u002Fantigravity-better-ui\u002Fenter-exit.md",{"path":82,"resourceType":69,"resourceTypeDesc":70,"sizeBytes":83,"sourceHtmlUrl":84},"icon-transitions.md",2647,"https:\u002F\u002Fgithub.com\u002Fboisenoise\u002Fskills-collections\u002Fblob\u002Fmain\u002Fskills\u002Fantigravity-better-ui\u002Ficon-transitions.md",{"path":86,"resourceType":69,"resourceTypeDesc":70,"sizeBytes":87,"sourceHtmlUrl":88},"icons.md",3963,"https:\u002F\u002Fgithub.com\u002Fboisenoise\u002Fskills-collections\u002Fblob\u002Fmain\u002Fskills\u002Fantigravity-better-ui\u002Ficons.md",{"path":90,"resourceType":90,"resourceTypeDesc":91,"sizeBytes":92,"sourceHtmlUrl":93},"LICENSE","许可证",1069,"https:\u002F\u002Fgithub.com\u002Fboisenoise\u002Fskills-collections\u002Fblob\u002Fmain\u002Fskills\u002Fantigravity-better-ui\u002FLICENSE",{"path":95,"resourceType":69,"resourceTypeDesc":70,"sizeBytes":96,"sourceHtmlUrl":97},"performance.md",2265,"https:\u002F\u002Fgithub.com\u002Fboisenoise\u002Fskills-collections\u002Fblob\u002Fmain\u002Fskills\u002Fantigravity-better-ui\u002Fperformance.md",{"path":99,"resourceType":69,"resourceTypeDesc":70,"sizeBytes":100,"sourceHtmlUrl":101},"surfaces.md",3584,"https:\u002F\u002Fgithub.com\u002Fboisenoise\u002Fskills-collections\u002Fblob\u002Fmain\u002Fskills\u002Fantigravity-better-ui\u002Fsurfaces.md",[103,112,144,174,189,205],{"skillNo":104,"name":105,"nameZh":6,"description":106,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":107,"licenseFileUrl":6,"sourceHtmlUrl":108,"categories":109,"keywords":110,"publishedAt":111},"01M742MX3Y2ZYK884ZXTRFKB9P","humanizer","Rewrite AI-sounding text so it reads like the writer without changing what it says.\nUse when editing or reviewing prose for AI tells: not-X-but-Y contrasts, one-line\nclosers, staged openers, forced triads, dashes everywhere, inflated claims, sales\nlanguage, stock AI words, bold labels, or filler.","MIT","https:\u002F\u002Fgithub.com\u002Fnodetool-ai\u002Fnodetool\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Fhumanizer\u002FSKILL.md",[],[],"2026-11-11 15:11:14",{"skillNo":113,"name":26,"nameZh":6,"description":114,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":115,"categories":116,"keywords":123,"publishedAt":39},"01M6BY4D93R0Q9400ZHJ04DXP1","Use the NodeTool MCP server to build, validate, run, and debug visual AI workflows, and to work with assets, collections, and jobs. Use when the user mentions NodeTool or asks for a workflow graph, a node, or a media generation run on their NodeTool server.","https:\u002F\u002Fgithub.com\u002Fnodetool-ai\u002Fnodetool\u002Fblob\u002Fmain\u002Fplugins\u002Fchatgpt\u002Fnodetool\u002Fskills\u002Fnodetool\u002FSKILL.md",[117,118],{"categoryNo":19,"categoryName":20,"taxonomyType":21,"taxonomyTypeDesc":22},{"categoryNo":119,"categoryName":120,"taxonomyType":121,"taxonomyTypeDesc":122},"01JBSEEDCAT000000000000030","开发工程师","PROFESSION","适用职业",[26,124,125,126,127,128,129,130,131,132,133,134,135,136,137,138,139,140,141,142,143,24],"MCP","Media","workflows","when","generation","media generation","with","visual","use","collections","for","node","work","debug","graph","validate","assets","run","server","mcp server",{"skillNo":145,"name":146,"nameZh":6,"description":147,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":148,"categories":149,"keywords":153,"publishedAt":39},"01M6BY4AEB05ZPBRVD078ATR2H","wan-2-6-prompting","Prompt Alibaba's Wan 2.6 video — text-to-video with a global style line plus timing-bracketed shots, and image-to-video that describes only what changes over time. Covers the shot-mode switch per provider, the prompt budget, negative prompts, motion intensity wording, audio input and prompt expansion. Use whenever the model id names a Wan 2.6 video route (alibaba\u002Fwan-2.6\u002Ftext-to-video, alibaba\u002Fwan-2.6\u002Fimage-to-video, wan-video\u002Fwan-2.6-t2v, wan-video\u002Fwan-2.6-i2v, wan-video\u002Fwan2.6-i2v-flash) on generate_video, animate_image or a TextToVideo node. Not for Wan 2.5 or 2.7, or the Wan 2.6 image routes.","https:\u002F\u002Fgithub.com\u002Fnodetool-ai\u002Fnodetool\u002Fblob\u002Fmain\u002Fpackages\u002Fsystem-skills\u002Fwan-2-6-prompting\u002FSKILL.md",[150],{"categoryNo":151,"categoryName":152,"taxonomyType":21,"taxonomyTypeDesc":22},"01JBSEEDCAT000000000000016","图像与多媒体",[154,155,156,157,158,159,160,135,161,162,163,164,165,166,167,168,169,170,171,172,173],"prompting","wan 2 6 prompting","line","expansion","input","image-to-video","text-to-video","route","only","mode","time","flash","Budget","generate","text","image","animate","model","Audio","Video",{"skillNo":175,"name":176,"nameZh":6,"description":177,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":178,"categories":179,"keywords":182,"publishedAt":39},"01M6BY47T4748ZG5ZMTAR97ZWZ","video-workflow","Build a reusable NodeTool video workflow for batches or repeated inputs. For a single finished piece, use a storyboard skill.","https:\u002F\u002Fgithub.com\u002Fnodetool-ai\u002Fnodetool\u002Fblob\u002Fmain\u002Fpackages\u002Fsystem-skills\u002Fvideo-workflow\u002FSKILL.md",[180,181],{"categoryNo":151,"categoryName":152,"taxonomyType":21,"taxonomyTypeDesc":22},{"categoryNo":19,"categoryName":20,"taxonomyType":21,"taxonomyTypeDesc":22},[24,183,173,184,185,186,187,132,188,26,134],"video workflow","reusable","storyboard","build","batches","single",{"skillNo":190,"name":191,"nameZh":6,"description":192,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":193,"categories":194,"keywords":196,"publishedAt":39},"01M6BY4587N2SYN315SXGEMZC3","video-clone","Analyze a reference video and recreate its shot structure with a new product or cast in NodeTool.","https:\u002F\u002Fgithub.com\u002Fnodetool-ai\u002Fnodetool\u002Fblob\u002Fmain\u002Fpackages\u002Fsystem-skills\u002Fvideo-clone\u002FSKILL.md",[195],{"categoryNo":151,"categoryName":152,"taxonomyType":21,"taxonomyTypeDesc":22},[197,198,173,130,199,200,26,201,202,203,204],"clone","video clone","cast","new","analyze","product","structure","reference",{"skillNo":206,"name":207,"nameZh":6,"description":208,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":209,"categories":210,"keywords":212,"publishedAt":39},"01M6BY4257C2KQYYMD9EK6DK8E","video-audio-continuity","Keep sound continuous across a multi-scene piece cut from generated video — why one clip per scene hard-cuts the audio at every boundary, when to write all the scenes into a single generation instead, the sound-design vocabulary that survives a provider's copyright filter, and how to check the length, the aspect ratio and the mix of what came back. Use when a piece has more than one scene and the video model writes its own audio (Seedance 2, Veo 3, MiniMax H3, Kling 2.6 and later), or when a cut's sound drops out at the shot changes.","https:\u002F\u002Fgithub.com\u002Fnodetool-ai\u002Fnodetool\u002Fblob\u002Fmain\u002Fpackages\u002Fsystem-skills\u002Fvideo-audio-continuity\u002FSKILL.md",[211],{"categoryNo":151,"categoryName":152,"taxonomyType":21,"taxonomyTypeDesc":22},[213,173,172,214,215,216,217,218,219,220,221,222,223,127,224,225,171,226,227,228,229],"video audio continuity","continuity","write","why","back","veo 3","at","cut","later","across","filter","generated","how to","mix","design","boundary","how"]