griptape-ai

Add per-row subtitle text and/or icons to an Options dropdown via update_ui_options, so each choice shows a name, a secondary description line, and optionally a Lucide icon.

Implement a Griptape Nodes Library node using SuccessFailureNode instead of ControlNode. Use whenever a node does something that can fail in a way a user might want to handle — file I/O, API calls, variable creation, JSON parsing, external services, or any operation where a partial failure is meaningful. SuccessFailureNode adds a failure output edge, a Status group with was_successful and result_details outputs, and re-raise behavior when no failure edge is wired.

Building and debugging custom parameter widgets for Griptape Nodes Library. Use when creating, fixing, or debugging JS widgets that display in node parameters — including widget lifecycle issues, value update flow, video/media display, URL resolution for the static file server, and the Python ↔ JS data bridge.

Hide or show node parameters by name at runtime — in __init__ for initial state, or in after_value_set to toggle visibility based on another parameter's value.

Add a ParameterString with an Options dropdown and a Button that refreshes the choices at runtime (e.g. querying a live API for available models, voices, etc.).

Implement a Griptape Nodes Library node for a proxied service from a spec file.

Add an info/warning/error badge to a Griptape Nodes parameter to surface docs links, examples, or contextual warnings in the UI.

Confirm a Griptape Node's parameter configurations against a live engine via MCP tools, guided by a survey document from the griptape-nodes-e2e-survey skill. Produces a structured markdown inspection report with live-confirmed parameter details for each configuration. The output root directory must be provided in the task prompt.

Propose a complete test plan for a Griptape Node, driven by the inspection report and survey document. The plan is a self-contained test specification — the only input the workflow skill needs. It includes parameter-value configurations, error test sections, helper nodes, MCP constraints, and everything else required to generate test workflows. Supports both creating new plans and editing existing ones. The output root directory must be provided in the task prompt.

Given a node class name, locate its source code via MCP tools and analyse it to enumerate all parameter configurations — value-driven, connection-driven, and UI-message-driven. Produces a survey document that guides the inspect skill's live exploration. The output root directory must be provided in the task prompt.

Reference wiki for planning Griptape Nodes test workflows. Use when deciding which nodes to create, how to wire them, and which assertion nodes to use to validate outputs — i.e. the planning phase that precedes code generation. Do NOT use for writing pytest code (use the griptape-nodes-e2e-sdk skill for that).

End-to-end test generation pipeline for a Griptape Node. Orchestrates the survey, inspect, plan, and workflow skills in sequence. Confirms the output directory with the user, manages subagent execution, presents the test plan for approval, and gates workflow generation on that approval.

Build, validate, and save test workflows for a Griptape Node via MCP tools, driven by the test plan. The plan is the sole input — it contains configuration tests, error tests, helper nodes, MCP constraints, and everything else needed to generate workflows. Produces one saved workflow per testable section. The output root directory must be provided in the task prompt.

Icon registry for the AnnotateImage widget — all available icon IDs, how to use mkIcon(), and how to add new icons. Load this when adding a new tool, button, or UI element that needs an icon.

Patterns for developing Python nodes in a Griptape Nodes library — base classes, parameter types, artifact handling, file output, lifecycle callbacks, and manifest registration. Load when adding or modifying nodes in this library.

Full layers system for the AnnotateImage node — data model, JS panel operations, imported layers, Python parity, and how to add new layer features. Load this when working on anything layer-related.

Architecture, data format, and capabilities of the AnnotateImage node and its JS widget. Use when reading, editing, or extending the annotate image node or its widget files.

Building and debugging custom parameter widgets for Griptape Nodes Library. Use when creating, fixing, or debugging JS widgets that display in node parameters — including widget lifecycle issues, value update flow, video/media display, URL resolution for the static file server, and the Python ↔ JS data bridge.

Create a Griptape Nodes library wrapping an open-source model repository. Takes an OS model repo URL and library repo path (already created from the template).

Set up a Griptape Nodes library structure for an OS model. Renames directories, adds the OS model as a git submodule, creates the advanced library file, and updates all configuration files.

Implement Griptape Nodes wrapping an OS model's features, based on a spec file. Creates node Python files, registers them in the manifest, and runs lint checks.

Research an open-source model repository and produce a structured specification for building a Griptape Nodes library.

Create a Griptape Nodes library wrapping an open-source model repository. Takes an OS model repo URL and library repo path (already created from the template).

Set up a Griptape Nodes library structure for an OS model. Renames directories, adds the OS model as a git submodule, creates the advanced library file, and updates all configuration files.

Implement Griptape Nodes wrapping an OS model's features, based on a spec file. Creates node Python files, registers them in the manifest, and runs lint checks.

Research an open-source model repository and produce a structured specification for building a Griptape Nodes library.

Build, run, and inspect Griptape Nodes workflows by driving the engine's MCP server. Use when the user asks to construct a node workflow, run an existing one, set parameter values, wire connections between nodes, or read output from a workflow run. Triggers include "build a workflow", "add a node to the flow", "connect these nodes", "run the flow", "what does node X output".

Build, run, and inspect Griptape Nodes workflows by driving the engine's MCP server. Use when the user asks to construct a node workflow, run an existing one, set parameter values, wire connections between nodes, or read output from a workflow run. Triggers include "build a workflow", "add a node to the flow", "connect these nodes", "run the flow", "what does node X output".

Reference guide for the inline {VAR} variable substitution system — architecture, data flow, key files, and extension points

Add a new diffusion model / pipeline TYPE to the modular_diffusion_nodes_library. Use ONLY when `<NewPipelineClass>.from_pipe(base_pipe)` cannot produce a working pipeline from the components already loaded on an existing base pipe — i.e. the new pipeline needs a component the base does not have, or needs differently-shaped/-trained weights for an existing component. Examples that qualify: Wan 2.2 T2V vs I2V (different UNet + image encoder), Qwen Edit (different transformer + image conditioning), Flux Fill (different transformer weights). Walks through the 5-step process: Provider enum, Standard Parameters, Runtime Parameters, Driver, Registration. DO NOT use when from_pipe(base) can build the new pipeline against the loaded components (ControlNet, inpaint, LTX media_gen_conditioning) — those are runtime variants and belong in add-pipeline-variants. Default to add-pipeline-variants first; only fall back to this skill when the from_pipe test fails.

Add a runtime variant (ControlNet, inpaint, or input-conditional pipe swap) to an EXISTING driver in modular_diffusion_nodes_library. DEFAULT TO THIS SKILL FIRST when adding any new capability to an existing model. The criterion is: can `<NewPipelineClass>.from_pipe(base_pipe)` produce a working pipeline from the components already loaded on the base pipe (UNet/transformer, VAE, text encoders, scheduler)? If yes — use this skill. Examples: ControlNet (any model), inpaint (any model), LTX media_gen_conditioning → LTXConditionPipeline. DO NOT use when from_pipe cannot produce a working pipeline (variant class needs a component the base does not have, or differently-shaped/-trained weights for an existing component) — that is a new pipeline type and belongs in add-modular-pipeline.

Author or update the user-facing documentation page for a node in modular_diffusion_nodes_library. Use whenever a new node is added, an existing node gains/loses/renames a parameter, its category or display name changes, its inputs/outputs change shape, or a provider/pipeline-type-specific behavior is added or removed. Also use when a docs page is missing a screenshot or has drifted from the node implementation. Produces or edits a page under docs/nodes/<name>.md following docs/node-doc-format.md, updates docs/index.md if the page is new, and flags missing screenshots. DO NOT use for code changes to the node itself — pair this with add-modular-pipeline or add-pipeline-variants when shipping a feature.

Set up a Griptape Nodes library structure for an OS model. Renames directories, adds the OS model as a git submodule, creates the advanced library file, and updates all configuration files.

Research an open-source model repository and produce a structured specification for building a Griptape Nodes library.