aws-agent-skills MCP — Cas d'usage, Installer & Démo en direct

Pourquoi l'utiliser

Fonctionnalités clés

18 core AWS services covered (IAM, Lambda, DynamoDB, S3, API Gateway, EC2, ECS, EKS, CFN, CloudWatch, RDS, SQS, SNS, Cognito, Step Functions, Secrets Manager, EventBridge, Bedrock)
Each SKILL.md includes CLI reference with real flags, not guessed ones
Common-pattern section per service surfaces production idioms
Security best-practices baked in (least-privilege IAM, encryption defaults)
Metadata tracks last-updated so you know the reference's freshness

Démo en direct

Aperçu en pratique

aws-agent-skill.replay ▶ prêt

0/0

Installer

Choisissez votre client

~/Library/Application Support/Claude/claude_desktop_config.json · Windows: %APPDATA%\Claude\claude_desktop_config.json

{
  "mcpServers": {
    "aws-agent-skill": {
      "command": "git",
      "args": [
        "clone",
        "https://github.com/itsmostafa/aws-agent-skills",
        "~/.claude/skills/aws-agent-skills"
      ],
      "_inferred": true
    }
  }
}

Ouvrez Claude Desktop → Settings → Developer → Edit Config. Redémarrez après avoir enregistré.

~/.cursor/mcp.json · .cursor/mcp.json

{
  "mcpServers": {
    "aws-agent-skill": {
      "command": "git",
      "args": [
        "clone",
        "https://github.com/itsmostafa/aws-agent-skills",
        "~/.claude/skills/aws-agent-skills"
      ],
      "_inferred": true
    }
  }
}

Cursor utilise le même schéma mcpServers que Claude Desktop. La config projet l'emporte sur la globale.

VS Code → Cline → MCP Servers → Edit

{
  "mcpServers": {
    "aws-agent-skill": {
      "command": "git",
      "args": [
        "clone",
        "https://github.com/itsmostafa/aws-agent-skills",
        "~/.claude/skills/aws-agent-skills"
      ],
      "_inferred": true
    }
  }
}

Cliquez sur l'icône MCP Servers dans la barre latérale Cline, puis "Edit Configuration".

~/.codeium/windsurf/mcp_config.json

{
  "mcpServers": {
    "aws-agent-skill": {
      "command": "git",
      "args": [
        "clone",
        "https://github.com/itsmostafa/aws-agent-skills",
        "~/.claude/skills/aws-agent-skills"
      ],
      "_inferred": true
    }
  }
}

Même format que Claude Desktop. Redémarrez Windsurf pour appliquer.

~/.continue/config.json

{
  "mcpServers": [
    {
      "name": "aws-agent-skill",
      "command": "git",
      "args": [
        "clone",
        "https://github.com/itsmostafa/aws-agent-skills",
        "~/.claude/skills/aws-agent-skills"
      ]
    }
  ]
}

Continue utilise un tableau d'objets serveur plutôt qu'une map.

~/.config/zed/settings.json

{
  "context_servers": {
    "aws-agent-skill": {
      "command": {
        "path": "git",
        "args": [
          "clone",
          "https://github.com/itsmostafa/aws-agent-skills",
          "~/.claude/skills/aws-agent-skills"
        ]
      }
    }
  }
}

Ajoutez dans context_servers. Zed recharge à chaud à la sauvegarde.

claude mcp add aws-agent-skill -- git clone https://github.com/itsmostafa/aws-agent-skills ~/.claude/skills/aws-agent-skills

Une seule ligne. Vérifiez avec claude mcp list. Supprimez avec claude mcp remove.

Cas d'usage

Usages concrets : aws-agent-skills

Write a least-privilege IAM policy for a Lambda + DynamoDB pair

👤 Backend engineers setting up AWS resources ⏱ ~15 min intermediate

Quand l'utiliser : You're wiring a Lambda to a Dynamo table and don't want to give it * permissions.

Déroulement

State the access needed

Use the aws iam + lambda skills. Write a policy for a Lambda that needs GetItem and UpdateItem on table Orders and nothing else.✓ Copié

→ Policy with exact actions + resource ARN, no wildcards
Test

Give me the aws iam simulate-custom-policy command to verify this allows only those two actions.✓ Copié

→ Runnable command with correct syntax

Résultat : A production-quality least-privilege IAM policy with verification.

Pièges

Policy uses wildcards for convenience — Insist on exact resource ARNs — skill supports this strongly

Combiner avec : terraform-skill

Design an EventBridge → Step Functions → Lambda pipeline

👤 Serverless engineers ⏱ ~40 min advanced

Quand l'utiliser : You need an event-driven workflow and want the right service boundaries.

Déroulement

Sketch the flow

Use the eventbridge + step-functions + lambda skills. Design a pipeline that triggers on S3 uploads, orchestrates 3 steps, and writes to DynamoDB.✓ Copié

→ Architecture diagram + IaC snippets
Observability

Now add CloudWatch alarms and structured logging.✓ Copié

→ Specific metric filters and alarm thresholds

Résultat : A buildable serverless architecture with monitoring.

Pièges

Step Functions costs explode at scale — Skill flags when Express workflows fit better than Standard

Combiner avec : terraform-skill

Scaffold a Bedrock-backed RAG service

👤 Engineers building LLM features on AWS ⏱ ~60 min advanced

Quand l'utiliser : You want to use Bedrock for embeddings + generation without piecing it together from blog posts.

Déroulement

Describe the service

Use the bedrock + s3 + dynamodb skills. Scaffold a RAG service: upload docs to S3, embed with Titan, store in DynamoDB, retrieve + generate at query time.✓ Copié

→ Runnable code with correct Bedrock invoke calls and embedding model IDs

Résultat : A first-pass RAG scaffold on AWS-native services.

Pièges

Using Dynamo for vector search at scale — Skill notes when OpenSearch or pgvector is a better fit

Combinaisons

Associez-le à d'autres MCPs pour un effet X10

aws-agent-skill + terraform-skill

AWS skills propose the design; Terraform skill writes the HCL

Design the Lambda+Dynamo stack using the aws skills, then have terraform-skill write the module.✓ Copié

aws-agent-skill + aws

Skill tells Claude the right calls; AWS MCP actually runs them

Use the aws-agent-skills to design the policy, then apply it via the aws MCP.✓ Copié

Outils

Ce que ce MCP expose

Outil	Entrées	Quand appeler
IAM skill	role/policy intent	Any AWS permissions work
Lambda skill	function spec	Serverless function design
DynamoDB skill	access patterns	NoSQL modeling
CloudFormation skill	resources	IaC in CFN
Bedrock skill	LLM use case	Using AWS-native GenAI

Coût et limites

Coût d'exécution

Quota d'API: None for the skill
Tokens par appel: 3-8k per service loaded
Monétaire: Free — skill is local. AWS usage costs are yours.
Astuce: Name the services you're using so Claude loads only those SKILL.md files.

Sécurité

Permissions, secrets, portée

Stockage des identifiants : No credentials in the skill. Real AWS calls should use OIDC roles, not static keys.

Sortie de données : None from the skill itself

Double-check generated IAM policies before deploying — the skill aims for least-privilege but you know your workload best.

Dépannage

Erreurs courantes et correctifs

CLI flag rejected by actual AWS CLI

AWS occasionally renames flags. Check the service's last-updated metadata; cross-reference official docs if stale.

Vérifier : aws <service> <command> help

Claude uses SDK v2 patterns for Python boto3

Boto3 is v1 (and stable); don't confuse with AWS SDK for JS v2/v3. Specify language explicitly.

Alternatives

aws-agent-skills vs autres

Alternative	Quand l'utiliser	Compromis
aws MCP	You want Claude to actually call AWS, not just write snippets	MCP executes; this skill informs
cloud-provider-agnostic IaC skills (pulumi, cdk)	You prefer code-first IaC	Different mental model

Plus

Ressources

📖 Lire le README officiel sur GitHub

🐙 Voir les issues ouvertes

🔍 Parcourir les 400+ serveurs MCP et Skills