aws-cli
Warn
Audited by Gen Agent Trust Hub on Nov 21, 2026
Risk Level: MEDIUMDATA_EXFILTRATIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [DATA_EXPOSURE_EXFILTRATION]: The skill uses commands that read and write to sensitive credential storage locations.
- Commands like
aws configure listandaws configureinteract directly with sensitive configuration data stored in~/.aws/credentialsand~/.aws/config. - The documentation explicitly identifies the paths for sensitive credential files, which can be leveraged for unauthorized access.
- [INDIRECT_PROMPT_INJECTION]: The skill processes data from external AWS services that could contain malicious instructions.
- Ingestion points: The skill reads data from AWS via
aws s3 ls,aws s3 cp, andaws lambda invoke(outputting tooutput.json). - Boundary markers: No delimiters or protective markers are used to separate untrusted external data from agent instructions.
- Capability inventory: The skill has broad capabilities including resource termination (
aws ec2 terminate-instances), data deletion (aws s3 rm), and IAM manipulation (aws iam create-access-key). - Sanitization: No validation or sanitization mechanisms are present to handle data retrieved from potentially attacker-controlled S3 buckets or Lambda outputs.
Audit Metadata