NVIDIA Corporation

Diagnose a specific failing, hanging, or incorrect Holoscan CLI command. The skill guides reproducible investigation, isolates the responsible layer, applies a minimal fix, and checks the same command and focused tests to verify the result.

• Diagnose a reproducible holoscan build failure
• Investigate a command that hangs
• Trace unexpected application output

Generate protein structures and sequences with specified conditions using NVIDIA's hosted API or a local NIM service. Supports binder and antibody design, reward-guided search, and inspection of generated candidates, outputs, and validation results.

• Generate protein candidates for a specified binding target
• Design antibodies or structures around a retained motif
• Explore candidates with reward-guided search

Create synthetic defect images for detector training using clean images and masks that mark defect regions. Guide mask placement, Cosmos fine-tuning, image generation, evaluation, quality refinement, and automatic assignment of provisional labels.

• Generating synthetic PCB defect images
• Placing defect masks on clean images
• Fine-tuning Cosmos for a defect dataset

Create and validate Cosmos Data Factory YAML configurations for changing video scenes, editing images, or generating videos. Run generation through remote model APIs, configure captions and quality checks, and retry candidates within a defined limit.

• Changing weather or lighting in source videos
• Editing selected attributes in still images
• Generating videos from an image or text

Guides CDF Auto-Labeling workflows for annotating data such as videos. Routes tasks to specialized references, covering shipped examples, stage selection, prompt and configuration changes, pipeline migration, container retrieval, and output validation.

• Starting automatic labeling for warehouse-safety videos
• Adapting a shipped workflow to a new domain
• Writing annotation prompts and question banks

Helps prepare video and image datasets with Cosmos Curator through splitting, filtering, deduplication, captioning, and embeddings, numerical representations for matching. Covers pipeline configuration, execution checks, Cosmos Dataset Search ingestion and search, and compatible handoffs to TAO data mining.

• Splitting and filtering videos into training datasets
• Deduplicating data and packaging it into archives
• Adding captions and embeddings to images

Diagnose GPU bugs in CUDA applications with NVIDIA Compute Sanitizer. Select the appropriate tool for memory access errors, shared-memory races, uninitialized reads, or synchronization faults, then build for source locations and interpret the diagnostic report.

• Locate illegal or out-of-bounds GPU memory access
• Investigate inconsistent results caused by shared-memory races
• Find reads of uninitialized GPU memory

Clarify an AI agent’s purpose, boundaries, and success criteria by reviewing its code and documents, then asking focused questions about intent. The resulting draft is handed to ethos-write to create an ethos.md record.

• Define the purpose and scope of a new agent
• Document the intended behavior of an existing agent
• Clarify priorities when desired outcomes conflict

Guide the creation and maintenance of an agent’s ethos.md, a record of its purpose, boundaries, and success criteria. Locate existing files and route requests to intent exploration or editing, carrying forward information already confirmed.

• Defining a new agent’s purpose and boundaries
• Locating an existing ethos.md and identifying gaps
• Updating specific rules or revisiting the agent’s intent

Turn an AI agent’s agreed purpose, behavior, boundaries, and success criteria into an ethos.md design document. The skill writes or edits the local file, validates its structure, and reviews the contents with the user.

• Document an agent design agreed through discussion
• Update an existing ethos.md when intent changes
• Record constraints and priorities for evaluation

Add optional NVIDIA Parabricks GPU steps to an existing genomics workflow while retaining its CPU path. Use runtime switches to compare outputs, execution time, and cost tradeoffs before enabling GPU processing for production.

• Identify genomics steps suitable for GPU acceleration
• Add switchable CPU and GPU paths to an existing workflow
• Compare CPU and GPU outputs and runtime

Prepare NVIDIA Parabricks commands for individual genomics tools, translate options from CPU tools, and check runtime readiness. Get guidance matched to the target release, including reference compatibility, command flags, and unresolved requirements before execution.

• Choosing a Parabricks tool for genomic data
• Translating STAR options into command drafts
• Checking GPU and container readiness

Set up Auto Ontology with Helm or Docker Compose, or connect an agent to an existing deployment through MCP, an agent connection protocol. Guides database connections, readiness checks, and troubleshooting for startup and sign-in problems.

• Installing Auto Ontology on Kubernetes
• Running the stack on one machine with Docker Compose
• Connecting an agent to an existing deployment

Guides the definition and publication of business terms, relationships, and measures in Auto Ontology. Covers SQL validation, model import and export, and persisted-change checks to verify that saved definitions reflect the intended business meaning.

• Define or revise business terms and relationships
• Check measure formulas, units, and aggregation levels
• Back up models and verify imports

Ask Auto Ontology questions in plain language, then check the generated SQL, returned data, and written answer. Supports MCP or authenticated REST access to verify that reported counts and other findings match the intended question.

• Checking a count against its SQL and returned data
• Confirming business terms before asking a question
• Reviewing incomplete or truncated query results

Guides fine-tuning public Encodon models on labeled protein-coding sequences or variants. Supports numerical prediction and classification, with data checks, LoRA adapters that train selected parameters, head-only training, or full training, followed by checkpoint and metric checks.

• Fine-tuning Encodon for sequence-level numerical prediction
• Training a classifier on labeled coding variants
• Preparing training data and a launch command

Prepare or run variant scoring with public CodonFM Encodon models. Validate coding sequences and codon positions in CSV inputs, assemble the scoring command, and check output arrays that compare reference and alternate codon probabilities in context.

• Validating a variant CSV before scoring
• Preparing an Encodon scoring command
• Running scoring and checking output arrays

Prepare a public CodonFM v1 environment and download supported Encodon checkpoints, the files containing trained model weights. Covers container or host setup, GPU access checks, and verification of downloaded weights and configuration files.

• Prepare a CodonFM development container
• Set up CodonFM directly on the host
• Download public Encodon model weights

Validate coding-sequence CSV files and extract embeddings, numerical features representing each sequence, with public CodonFM/Encodon models. Prepare extraction commands, choose suitable checkpoints, and verify outputs for downstream modeling of translation efficiency, expression, or mRNA stability.

• Validate coding-sequence CSV files before extraction
• Extract numerical sequence features with Encodon
• Choose a checkpoint for downstream property modeling

Turn records from a stopped NVFLARE Auto-FL optimization campaign into a reproducible Markdown report and JSON summary. Summarize improvements, retained candidates, failed ideas, and literature outcomes, with a refreshed progress plot when plotting dependencies are available.

• Create a final report after an optimization campaign stops
• Summarize baseline scores and retained improvements
• Review successful and unsuccessful literature-based ideas

Fine-tune the MAISI VAE, a model that compresses and reconstructs CT or MRI volumes, using NIfTI datasets. Prepare training settings, check inputs before training, and save model checkpoints and loss histories for review.

• Check a NIfTI dataset before GPU training
• Fine-tune a VAE with CT or MRI volumes
• Compare reconstruction losses across training runs

Segment specified organs in CT NIfTI scans using NV-Segment-CT VISTA3D. Produces a label map identifying organ regions and JSON evidence covering labels, voxel counts, physical volumes, geometry checks, and model identity for engineering validation.

• Segmenting the liver, spleen, and both kidneys in CT scans
• Recording voxel counts and physical volumes for each organ
• Checking output geometry and unexpected labels

Run NV-Reason-CT on a supplied 3D chest or abdominal CT scan and a text prompt for research or engineering. Returns generated text and JSON records of input geometry, model identity, dependencies, and limitations.

• Testing inference on a chest CT scan
• Running abdominal CT inference with a focused question
• Checking local model assets and runtime readiness

Run NV-Segment-CTMR to divide CT or MRI NIfTI volumes into labeled regions. Save the label map and records of region volumes, processing time, and geometry checks for technical verification, without clinical interpretation.

• Automatically segmenting CT body volumes
• Segmenting body MRI or preprocessed brain MRI
• Requesting regions by specific label IDs