aws
Operate AWS accounts and SSO profiles with aws and aws-sso-util.
Install
npx skills add https://github.com/fmind/dot/tree/main/skills/aws
claude plugin marketplace add https://llmmart.ai/marketplace.json && claude plugin install fmind-dot@llmmart
git clone https://github.com/fmind/dot.git
The skills CLI installs just this skill, for any of its supported agents. Claude Code installs the whole fmind/dot collection as a plugin from our marketplace. Git is the plain clone.
Skill manifest
Amazon Web Services CLI
Operate AWS accounts, IAM, S3, ECS, and CloudWatch with aws.
Workflow
Verify profile and caller identity: never assume roles or run under ambiguous profiles:
aws sts get-caller-identity --profile <profile> --output json --no-cli-pager aws configure list-profilesSign in with short-lived credentials: prefer IAM Identity Center (
aws sso login), falling back toaws login(--remoteon headless hosts) instead of static access keys:aws sso login --profile <profile> aws login --profile <profile>Pin
--profileand--regionon every consequential call so env vars cannot redirect targets.Bound reads first: combine server-side filters,
--max-items, and--query(JMESPath) before mutating state:aws s3 ls --profile <profile> --region <region> --no-cli-pager aws ecs list-clusters --profile <profile> --region <region> --max-items 20 --output json --no-cli-pagerAuthorize mutations: resource creation, security groups, IAM policies, and deletions require user approval (reusing existing task authority).
Gotchas
--queryfilters client-side: pair with--max-items(--page-sizeonly sizes HTTP pages), keepNextTokenin projections, and flag capped output as partial. Pass--no-cli-pager --cli-error-format jsonfor structuredCode/Messagediagnostics without credential-leaking--debug.- Failures are findings: report
AccessDeniedExceptionor missing roles directly instead of escalating IAM permissions.
Official Skills
Upstream: AWS Agent Toolkit. Inspect read-only via aws agent-toolkit list-available-skills / search-skills --search-query <topic>, get-skill-metadata --skill-name <name>, and get-skill-file --skill-name <name> --file-path SKILL.md --skill-version <version>, installing only under vendor-skill policy. Never run aws configure agent-toolkit --yes (bulk-installs skills and MCP across hosts; AWS_CLI_AGENT_TOOLKIT_HINT_DISABLED=true suppresses hints).
Documentation
- AWS CLI User Guide · AWS CLI Command Reference
- Releases: AWS CLI GitHub Releases
- Companion skills: infra-as-code (provisioned infrastructure), code-security (IAM audits).
Files (dot)
-
SKILL.md 2.8 KB
--- name: aws description: "Operate AWS accounts and SSO profiles with the aws CLI." license: MIT metadata: kind: connector author: Médéric HURIER (Fmind) source: github.com/fmind/dot/tree/main/skills/aws created: "2026-09-16" updated: "2026-11-10" --- # Amazon Web Services CLI Operate AWS accounts, IAM, S3, ECS, and CloudWatch with `aws`. ## Workflow 1. **Verify profile and caller identity**: never assume roles or run under ambiguous profiles: ```bash aws sts get-caller-identity --profile <profile> --output json --no-cli-pager aws configure list-profiles ``` 1. **Sign in with short-lived credentials**: prefer IAM Identity Center (`aws sso login`), falling back to `aws login` (`--remote` on headless hosts) instead of static access keys: ```bash aws sso login --profile <profile> aws login --profile <profile> ``` 1. **Pin `--profile` and `--region`** on every consequential call so env vars cannot redirect targets. 1. **Bound reads first**: combine server-side filters, `--max-items`, and `--query` (JMESPath) before mutating state: ```bash aws s3 ls --profile <profile> --region <region> --no-cli-pager aws ecs list-clusters --profile <profile> --region <region> --max-items 20 --output json --no-cli-pager ``` 1. **Authorize mutations**: resource creation, security groups, IAM policies, and deletions require user approval (reusing existing task authority). ## Gotchas - **`--query` filters client-side**: pair with `--max-items` (`--page-size` only sizes HTTP pages), keep `NextToken` in projections, and flag capped output as partial. Pass `--no-cli-pager --cli-error-format json` for structured `Code`/`Message` diagnostics without credential-leaking `--debug`. - **Failures are findings**: report `AccessDeniedException` or missing roles directly instead of escalating IAM permissions. ## Official Skills Upstream: [AWS Agent Toolkit](https://docs.aws.amazon.com/agent-toolkit/latest/userguide/). Inspect read-only via `aws agent-toolkit list-available-skills` / `search-skills --search-query <topic>`, `get-skill-metadata --skill-name <name>`, and `get-skill-file --skill-name <name> --file-path SKILL.md --skill-version <version>`, installing only under [vendor-skill policy](../agent-project/references/vendor-skills.md). Never run `aws configure agent-toolkit --yes` (bulk-installs skills and MCP across hosts; `AWS_CLI_AGENT_TOOLKIT_HINT_DISABLED=true` suppresses hints). ## Documentation - [AWS CLI User Guide](https://docs.aws.amazon.com/cli/latest/userguide/) · [AWS CLI Command Reference](https://docs.aws.amazon.com/cli/latest/) - Releases: [AWS CLI GitHub Releases](https://github.com/aws/aws-cli/releases) - Companion skills: [infra-as-code](../infra-as-code/SKILL.md) (provisioned infrastructure), [code-security](../code-security/SKILL.md) (IAM audits).
Comments (0)
Sign in to join the conversation.
Reviews (0)
No reviews yet.
No comments yet.