CatalogueSecurity & Compliance
Cloud Permission & IAM Least-Privilege Audit
Reads cloud permissions for access that is wider than the job, public by accident, or left on a long-lived key.
What it examines
- Wildcard actions and resources where a narrower set would do
- Trust policies that let too many principals assume a role
- Public buckets, snapshots, and images
- Security groups open to the world, especially on admin and database ports
- Volumes, buckets, and databases without encryption at rest
- Missing CloudTrail, flow logs, or access logs
- Long-lived access keys where a role would do
What you get back
A findings file. Each item has the evidence, how serious it is, and a fix.
Dependencies
- An AI coding agent that reads SKILL.md files, such as Claude Code, Codex or Gemini CLI
- Access to the AWS account through a read-only IAM role. The audit never asks for write access.
- A way for the agent to read IAM, security group, storage and logging settings with that role, such as the AWS CLI signed in to it
- Optional but strongly preferred: CloudTrail history. Without it, tightening advice is marked lower-confidence.
- Optional: the infrastructure code (Terraform, CloudFormation or CDK), to spot drift between what is declared and what is deployed
- Python 3 with the jsonschema package, to check findings.json before fixing anything
How it works
The audit is read-only. It lists IAM roles, users and policies, then compares what each one is allowed to do with what CloudTrail shows it actually did.
It checks seven things: wildcard permissions, over-broad trust policies, public exposure, security groups open to the internet, unencrypted storage, disabled logging and long-lived access keys.
A permission is only called removable when it was granted but never used in the history it saw. You get a ranked report and staged policy proposals, never policies ready to apply.
Setup
- Unzip the download. Copy the audit and remediate folders into your agent's skills folder (for Claude Code, .claude/skills/) in the repository you want to check.
- Sign in to AWS with a read-only role: run aws sso login, or set up a profile with aws configure, then confirm with aws sts get-caller-identity.
- If you have CloudTrail history or the infrastructure code, keep them ready. The audit asks for them.
- Open the repository in your agent and ask it to run the a5-cloud-permission-iam-audit skill. Tell it which profile and region to use.
- Check the result with the validator: pip install jsonschema, then python3 schema/validate.py findings.json.
- Ask your agent to run the a5-cloud-permission-iam-remediate skill using findings.json. Fix some findings and leave the rest open if you like. Test any tightened policy for a while before it goes to production.
Run it with your agent
Copy the prompt, fill in the parts in square brackets, and paste it into your agent.
Run the a5-cloud-permission-iam-audit skill on this repository. Use the AWS profile [profile name] in region [region]. It is a read-only role, so do not ask for write access. CloudTrail history: [available, or not]. Infrastructure code: [path, or none]. Mode: audit only. Do not change any code. Write findings.json and findings.md in the repository root.
Run the a5-cloud-permission-iam-remediate skill using findings.json. Work through the findings I choose, and leave the rest open.