from agent-plugins542
Generates deployment code and a Jupyter notebook to deploy LoRA fine-tuned Nova or OSS models (from SageMaker Serverless Model Customization) to SageMaker endpo
This skill determines the correct deployment pathway for models fine-tuned via SageMaker Serverless Model Customization and generates the deployment artifacts (including a Jupyter notebook and sample invoke code). It inspects training job metadata, extracts model IDs and tags, identifies whether the model is Nova or OSS, and guides the user through choosing a target (SageMaker endpoint vs Bedrock). The skill validates model eligibility, displays license terms, runs pathway-specific steps, and produces a post-deployment summary with usage examples and cleanup instructions.
Use this skill when a user asks to "deploy my model", "create an endpoint", or wants guidance on making a fine-tuned model available in production. It's specifically for LoRA-fine-tuned models created with SageMaker Serverless Model Customization (not full-fine-tuned models or base models).
Designed for coding/ops-capable agents that can call AWS tools and run scripts (e.g., agents with CLI/tool integrations like Copilot/Code agents).
This skill guides deploying LoRA fine-tuned Nova/OSS models from SageMaker Serverless Model Customization to SageMaker endpoints or Bedrock. The SKILL.md is well-structured with clear decision trees, but the 4 bundled scripts are Jupyter notebook cells (using %pip magic) and fail as standalone Python. No security concerns — scripts use standard AWS SDK calls with placeholder config variables.
Scripts are notebook-format (.ipynb cell content) run as .py — the %pip magic syntax is the only failure cause. If executed in Jupyter they would likely work. SKILL.md explicitly states notebooks should be written via file write or MCP tools, not bash — the scripts here appear to be reference notebook cells packaged as .py files.
