The Agent Skills Directory

[SAFE]: The skill provides legitimate security remediation steps for AWS S3 buckets. The inclusion of a Python script using boto3 is a standard approach for cloud auditing, and the use of placeholder account IDs and ARNs is consistent with safe documentation practices.
[COMMAND_EXECUTION]: The SKILL.md file provides multiple AWS CLI commands for identifying and remediating bucket misconfigurations.
Evidence: Commands such as aws s3api list-buckets, aws s3control put-public-access-block, and aws organizations create-policy are provided to manage cloud resources as part of the skill's intended purpose.
[INDIRECT_PROMPT_INJECTION]: The skill possesses a vulnerability surface for indirect prompt injection as it processes untrusted metadata from an external AWS account (e.g., bucket names, policies, and ACLs).
Ingestion points: The scripts/agent.py script retrieves bucket metadata using list_buckets(), get_bucket_policy(), and get_bucket_acl().
Boundary markers: No explicit delimiters are mentioned; however, the script utilizes programmatic JSON parsing and specific field checks rather than raw prompt interpolation.
Capability inventory: The script has the capability to modify AWS resource settings via put_public_access_block() and put_bucket_encryption().
Sanitization: The script performs structural validation of policies (e.g., parsing JSON and identifying specific Principal values), which mitigates basic injection attempts within the metadata.

remediating-s3-bucket-misconfiguration