AWS Agent Skills
A curated skill pack that gives Claude Code and Codex deep AWS expertise across 18 core services for automated cloud engineering.
Evidence shows the repository lacks permission models, user confirmation mechanisms, data flow transparency, sensitive data handling policies, dependency security audits, external effects documentation, rollback plans, or source attribution. All trust criteria are absent, hence score 0.
Self-consistency: README claims 18 services but mentions rlhf skill in examples, not in the list, inconsistency, score 1. Dependency availability: relies on GitHub Actions and Python packages, but requirements.txt content not provided, cannot verify, score 1. Failure messages: no error handling or failure documentation, score 0.
Audience and scenarios: README clearly targets Claude Code and Codex users with usage examples, score 2. Capability boundaries: no explicit limitations or boundaries, score 1. Trigger precision: SKILL.md template includes trigger phrases but no actual skill files, score 1. Environment fit: multiple installation methods but no compatibility notes for versions, score 1.
Information architecture: README well-structured with installation, skill list, usage examples, score 2. Install notes: multiple installation methods provided, score 2. Naming stability: skill names align with AWS services but no version stability, score 1. Examples and FAQ: usage examples provided but no FAQ, score 2. Known limitations: none mentioned, score 0. License: MIT license file present, score 2. Versioning and changelog: no version or changelog, score 0. Maintenance responsibility: contribution guide and automated update workflow but no explicit maintainer, score 1.
Output usability: skill content not provided, cannot assess output quality, score 1. Marginal value: claims LLM-optimized knowledge base but no concrete content, score 2. Cost-benefit: claims token efficiency but no data, score 1.
Claim traceability: README claims not linked to specific files or evidence, score 1. Cross-source corroboration: no external sources, score 0. Fact-inference separation: README distinguishes facts and inferences but no specific basis, score 1.
- Repository lacks skill content files, actual functionality cannot be verified.
- README mentions rlhf skill not in the list, inconsistency.
- No dependency security audit or permission model, use with caution.
What does this agent do, and when should you use it?
AWS Agent Skills is a plugin/skill collection for Claude Code and Codex, covering 18 AWS services like IAM, Lambda, DynamoDB, S3, and more. Unlike AWS MCP, it is reasoning-first: it provides a local, LLM-optimized knowledge base with real-world patterns, edge cases, and best practices, rather than streaming live docs or schemas. Skills are auto-checked against AWS documentation on a weekly basis to stay current. Each skill includes a SKILL.md file with core concepts, CLI references, and troubleshooting, plus supplementary files. Installation varies by client: Claude Code uses plugin marketplace or direct install; Codex uses the skill-installer command.
Once installed, users can ask Claude Code or Codex natural-language questions about AWS tasks, e.g., "Create an IAM policy for Lambda to access DynamoDB". The skill pack provides structured knowledge (SKILL.md and supplementary files) to guide the agent in generating IaC templates, debugging errors, and offering best practices. It does not make direct AWS API calls; instead, it empowers the agent's reasoning to output configurations, code, or commands. The install flow uses /plugin marketplace add and /plugin install for Claude Code, and $skill-installer install for Codex.
- A developer using Claude Code needs to quickly generate an IAM policy for Lambda to access DynamoDB.
- A DevOps engineer debugging Lambda timeout issues in Claude Code, with advice on VPC setup.
- A Codex CLI user needs a CloudFormation template for a serverless API.
- A developer creating an ECS Fargate service with a load balancer.
- A platform engineer setting up EventBridge rules for scheduled tasks.
What are this agent's strengths and limitations?
- Covers 18 core AWS services, providing broad cloud engineering knowledge.
- Localized, compressed knowledge base saves tokens compared to MCP, keeping context windows small and predictable.
- Weekly automatic checks against AWS documentation ensure skills stay current with service changes.
- Works with both Claude Code and Codex, with straightforward integration.
- Only supports Claude Code and Codex; other AI agents are not supported.
- Does not directly call AWS APIs, relying on the agent to generate commands or code, which may be less direct than MCP.
- Installation requires different steps per client, and Codex requires an external skill-installer tool.
- Skill maintenance depends on the author's continued updates; users must actively update to get new content.
How do you install or deploy this agent?
For Claude Code:
- From Marketplace: run /plugin marketplace add itsmostafa/aws-agent-skills, then /plugin install aws-agent-skills.
- From GitHub: run /plugin install https://github.com/itsmostafa/aws-agent-skills.
- For local development: run /plugin install ./path/to/aws-agent-skills.
For Codex CLI:
- Run $skill-installer install https://github.com/itsmostafa/aws-agent-skills/<skill-name>, e.g., $skill-installer install https://github.com/itsmostafa/aws-agent-skills/rlhf. Requires Codex CLI and the skill-installer tool.
How do you use this agent?
After installation, within Claude Code, simply ask in natural language, e.g., "Create an IAM policy for Lambda to access DynamoDB" or "Debug this access denied error". In Codex, load the relevant skill and similarly ask. The agent will use the hidden knowledge to generate configurations, code, or commands.
How does this agent compare with similar options?
Compared to AWS MCP, AWS Agent Skills focuses on reasoning-first, providing local knowledge without streaming large docs or schemas, offering better token efficiency and avoiding MCP infrastructure and credential exposure.