awslabs/agent-plugins
33 skills2.9k installs28.1k starsGitHub
Install
npx skills add https://github.com/awslabs/agent-pluginsSkills in this repo
1Aws LambdaDesign, build, deploy, test, and debug serverless applications with AWS Lambda. Triggers on phrases like: Lambda function, event source, serverless application, API Gateway, EventBridge, Step Functions, serverless API, event-driven architecture, Lambda trigger. For deploying non-serverless apps to AWS, use deploy-on-aws plugin instead. Design, build, deploy, and debug serverless applications with AWS serverless services. This skill provides access to serverless development guidance through the AWS Serverless MCP Server, helping you to build production-ready serverless applications with best practices built-in.222installs2DeployDeploy on AWS is a skill that takes an existing application and deploys it to AWS with minimal user decisions. A developer uses it to scan a codebase, get an AWS service recommendation with a monthly cost estimate, and generate infrastructure-as-code with security defaults applied. It runs IaC security scans before executing the deployment on user confirmation.205installs3Aws Architecture DiagramThis skill generates AWS architecture diagrams as draw.io XML files using official AWS4 icons, matching the style of AWS Reference Architecture diagrams. A developer uses it to visualize existing infrastructure from CloudFormation, CDK, or Terraform code, or to brainstorm an architecture from scratch. It validates the XML and can export to PNG, SVG, or PDF via the draw.io CLI.174installs4Directory ManagementManages project directory setup and artifact organization. Use when starting a new project, resuming an existing one, or when a PLAN.md needs to be associated with a project directory. Creates the project folder structure (specs/, scripts/, notebooks/, manifests/, agent_memory/) and resolves project naming. Before any work begins, resolve the project name:147installs5Use Case SpecificationAWS agent plugin skill for drafting use case specification files recommended before model customization. Captures the business problem statement, identified stakeholders, success metrics, and constraints in a structured document reusable across customization pipelines. Designed to pair with the planning skill which must activate alongside other customization skills. Output becomes the authoritative input for downstream training, fine-tuning, or Bedrock customization steps, ensuring measurable criteria before infrastructure or data prep work begins.138installs6PlanningAWS agent plugin planning skill that must activate alongside any other customization skill. Discovers user intent through clarifying questions, maps the customization goal to AWS recommended workflow stages, and outputs a structured step-by-step plan before data prep, training, or deployment skills execute. Ensures customization projects start with explicit scope, constraints, and sequencing rather than jumping directly into infrastructure or dataset tasks. Serves as the mandatory planning gate for the AWS model customization plugin family.133installs7Aws Serverless DeploymentAWS serverless deployment skill covering SAM and CDK paths with reference files for project setup, Lambda constructs, serverless patterns, and coexistence. SAM guidance includes sam_init templates, Globals defaults, samconfig.toml environments, and container builds. CDK guidance favors TypeScript L2 constructs, grant methods, stack separation for stateful resources, cdk.context.json commits, assertions tests, and cdk diff in CI. Requires AWS CLI credentials, SAM CLI, Docker for local invoke, and optional AWS Serverless MCP with write access. Includes PostToolUse sam validate hook on template edits. Default framework is CDK with override syntax. Routes Lambda runtime behavior to the aws-lambda skill and general app deployment to deploy-on-aws.130installs8Dataset EvaluationDataset evaluation is a skill that validates dataset formatting and quality for Amazon SageMaker model fine-tuning and evaluation. A developer uses it before starting a fine-tuning job to detect the file format and check schema compliance against the chosen model family and technique (SFT, DPO, or RLVR). It runs a bundled format_detector.py script and reports whether the data is ready for training or evaluation.121installs9FinetuningFinetuning is a skill that generates code to fine-tune a base model using Amazon SageMaker serverless training jobs. A developer uses it once a technique and base model are selected and a dataset is uploaded, to produce training code for SFT, DPO, RLVR, or RLAIF. It handles RLVR reward functions, RLAIF custom prompts, and EULA review, following strict template-based code generation.121installs10Model DeploymentThis skill generates code to deploy fine-tuned models from SageMaker Serverless Model Customization to a SageMaker endpoint or Bedrock. It identifies the model type (Nova vs OSS), lets the user choose a deployment target, displays the license, and generates the deployment code. A developer uses it when they want to make a LoRA fine-tuned model available for inference.121installs11Dataset TransformationDataset transformation is a skill that generates code to convert datasets between machine-learning schemas for model training or evaluation. A developer uses it when a dataset's format must change to match a target training technique or evaluation format instead of writing inline conversion code. It supports formats including OpenAI chat, SageMaker SFT/DPO/RLVR/RLAIF, HuggingFace preference, and Bedrock Nova, and follows a step-by-step confirm-before-proceeding workflow.120installs12Hyperpod SsmThis skill runs remote commands and transfers files on SageMaker HyperPod cluster nodes through AWS Systems Manager, which is the primary access path because direct SSH is not available. It bundles scripts to resolve cluster info, list nodes, and execute or transfer files per node, and other HyperPod skills depend on it for node-level operations. A developer uses it whenever a task needs shell access to HyperPod instances.117installs13Hyperpod Version CheckerThis skill checks and compares software component versions on SageMaker HyperPod cluster nodes, including NVIDIA drivers, CUDA, cuDNN, NCCL, EFA, MPI, Neuron SDK, Python, and PyTorch. It runs a bundled script that analyzes CUDA/driver compatibility and, with JSON output, makes cross-node diffing easy. A developer uses it to detect version mismatches, plan upgrades, or document cluster configuration.117installs14Hyperpod Issue ReportThis skill gathers diagnostic logs and configuration from HyperPod cluster nodes over SSM and stores the results in S3, supporting both EKS and Slurm with auto-detection. It runs a bundled hyperpod_issue_report.py for parallel collection across nodes, then reports the S3 location and offers analysis. A developer uses it to document cluster state or prepare an AWS Support case.115installs15Model EvaluationThis skill generates Python code to evaluate SageMaker models, both base and fine-tuned, using either LLM-as-Judge or a Custom Scorer. It determines the evaluation type, validates it against the model type and whether an eval dataset exists, then hands off to the matching evaluation workflow. A developer uses it to benchmark a model, test its performance, or compare models.96installs16Amplify WorkflowThis skill builds and deploys full-stack web and mobile apps with AWS Amplify Gen2 using a TypeScript code-first approach. A developer uses it to define backend resources like Cognito auth, AppSync/DynamoDB data, and S3 storage, then integrate a frontend across eight supported frameworks. It routes to version-specific reference files to prevent common Amplify Gen2 mistakes.84installs17Aws Lambda Durable FunctionsAWS Lambda durable functions is a skill for building long-running, multi-step serverless applications on AWS Lambda with automatic state persistence and retry logic. A developer uses it when writing stateful Lambda handlers that must survive interruptions, coordinate steps, wait on callbacks, or implement saga-style compensation. It codifies the replay model rules and provides code patterns, IaC deployment guidance, and local testing setup.71installs18Amazon Location ServiceThis skill integrates Amazon Location Service geospatial APIs into AWS applications, covering maps, geocoding, routing, places search, geofencing, and device tracking. A developer uses it to add location features to a web or mobile app, defaulting to the bundled JavaScript client and resourceless API operations. It documents authentication choices, API selection guidance, and common integration mistakes.69installs19Hyperpod Performance DebuggerThis skill diagnoses two performance problems on Amazon SageMaker HyperPod clusters: uneven NCCL bandwidth across nodes and poor filesystem throughput. Its read-only perf-snapshot.sh gathers host-side signals such as Xid, ECC, NVLink, EFA reachability, and FSx saturation, then routes each concern to the owning sibling skill. A developer uses it when training is slow due to a straggler node or a saturated filesystem.65installs20Hyperpod NcclThis skill diagnoses NCCL and adjacent training-pod failures on SageMaker HyperPod GPU clusters, including training hangs, AllReduce timeouts, EFA errors, container and GPU OOM, and CrashLoopBackOff. Its read-only nccl-diagnose.sh reads state via AWS APIs, kubectl, and SSM and points each finding at a remediation section. A developer uses it when distributed GPU training stalls or collective ops time out.64installs21Hyperpod Node DebuggerThis skill diagnoses and remediates per-node problems on a HyperPod cluster running EKS or Slurm, covering EFA, GPU hardware faults (XID, ECC, NVLink, row-remap, DCGM), disk and memory pressure, and lifecycle-script failures. Its read-only triage-cluster.sh collects node state and points each finding at a reference section. A developer uses it when one node is unhealthy, unresponsive, or needs replacing.64installs22Hyperpod Slurm DebuggerThis skill diagnoses Slurm scheduler and node-daemon issues on Amazon SageMaker HyperPod Slurm clusters, such as nodes stuck in down or drain, jobs stuck PENDING or COMPLETING, wrong GPU counts, and unresponsive slurmctld. It runs a diagnostic script, relays output verbatim, and maps each finding to official AWS or Slurm docs without printing any state-mutating command. A developer uses it to triage a stuck Slurm node or job on HyperPod.64installs23Hyperpod Cluster DebuggerThis skill diagnoses cluster-wide problems on Amazon SageMaker HyperPod clusters running EKS or Slurm, such as CloudFormation failures, EFA health-check failures, capacity errors, and dangling nodes. It runs a read-only diagnose-cluster.sh that prints each issue as a [FAIL] pointing to a remediation section, and it never changes cluster state. A developer uses it when a HyperPod cluster fails to create or behaves incorrectly.63installs24Api GatewayThis skill provides guidance for building, governing, and operating APIs with Amazon API Gateway across REST, HTTP, and WebSocket types. A developer uses it to choose the right API type, configure authorizers, usage plans, throttling, custom domains, and CORS, or troubleshoot gateway errors. It supplies working SAM/CloudFormation templates from bundled reference files.54installs25Model SelectionModel-selection guides a developer through picking a base foundation model by querying SageMaker Hub in the user's AWS region. It lists available models with their licenses and benchmark data, then confirms a choice. A developer uses it when asked which model to use or when a specific model family like Llama, Mistral, or Nova needs its exact Hub model ID resolved.42installs26Sdk Getting StartedSdk-getting-started validates that the user's environment can run Amazon SageMaker AI operations. It checks the SageMaker Python SDK version, the AWS region, and the execution role, resolving them programmatically rather than asking the user. A developer runs it first, before any SageMaker or Bedrock training, evaluation, or deployment work.40installs27Finetuning TechniqueThis skill helps a developer pick a fine-tuning technique (SFT, DPO, RLVR, or RLAIF) for a use case and checks that the selected model actually supports it on SageMaker. It reads a use-case spec, recommends a technique, then runs get_recipes.py against the model and hub to confirm availability. A developer uses it after deciding to fine-tune and before running training.39installs28DsqlDSQL is a skill for building applications with Amazon Aurora DSQL, a serverless, PostgreSQL-compatible distributed SQL database. A developer uses it to manage schemas, run queries and migrations, diagnose query plans, handle IAM auth and multi-tenant patterns, and load data in bulk. It covers converting MySQL and PostgreSQL schemas to DSQL, replacing foreign keys, and implementing optimistic-concurrency-control retry logic.34installs29Document ServiceDocument Service is a skill that analyzes an existing codebase to produce structured technical documentation and architecture diagrams with source-of-truth citations. A developer uses it when inheriting an undocumented service and needing to understand or onboard onto it, especially for AWS workloads built with CDK, CloudFormation, or Terraform. Every finding links back to the exact file and line it was derived from.22installs30Aws TransformAWS Transform is a skill for migrating, modernizing, and upgrading existing codebases to AWS. A developer uses it to port .NET Framework to .NET 8/10, convert mainframe COBOL to Java, rehost VMware VMs to EC2, migrate SQL Server to Aurora, or upgrade language and SDK versions. It also runs continuous-modernization analysis over repositories to find tech debt, CVEs, and security vulnerabilities and route each workload to the right transformation.21installs31Elastic BeanstalkElastic Beanstalk is a skill for deploying web and worker applications to production on AWS Elastic Beanstalk, where the developer provides code and AWS manages deployment, scaling, patching, and health. A developer uses it when they want a managed application platform (a Heroku-like experience) rather than serverless or hand-managed container orchestration. It maps the app to a platform branch, configures the environment, and generates CLI, CDK, or Terraform.13installs32Aws Step FunctionsAWS Step Functions is a skill for building serverless workflows as state machines using Amazon States Language (ASL) with the JSONata query language. A developer uses it to author, validate, and test workflows that orchestrate microservices, automate processes, and build data or ML pipelines. It covers state types, error handling, AWS service integrations, and migrating existing state machines from JSONPath to JSONata.11installs33Aws Lambda Managed InstancesAWS Lambda Managed Instances is a skill for evaluating, configuring, and migrating workloads to AWS Lambda Managed Instances (LMI), which run Lambda code on EC2 instances that AWS manages. A developer uses it when deciding whether steady high-volume traffic should move off standard Lambda for cost or cold-start reasons, and when sizing instances and checking concurrency safety. It builds a required cost comparison before recommending LMI.9installs