The Agent Skills Directory

[DATA_EXFILTRATION]: The skill demonstrates commands for reading sensitive local files, such as ~/.ssh/id_rsa, to upload them to AWS Secrets Manager. While this is a legitimate use case for the skill, accessing private keys is a sensitive operation that requires careful handling.
[INDIRECT_PROMPT_INJECTION]: The skill's audit and reporting scripts process data retrieved from AWS Secrets Manager and RDS APIs, creating a potential surface for indirect prompt injection if an attacker can control the metadata or content of the stored secrets.
Ingestion points: Secret list data fetched via client.list_secrets() and the aws secretsmanager list-secrets CLI command.
Boundary markers: The scripts do not use specific delimiters or instructions to prevent the agent from interpreting data within the secret fields as instructions.
Capability inventory: The skill possesses the capability to execute shell commands, perform network requests via the requests library, and interact with databases using pymysql.
Sanitization: The scripts rely on standard JSON parsing and do not implement additional sanitization or validation of the secret content.
[COMMAND_EXECUTION]: The skill heavily utilizes the AWS CLI and custom scripts to manage cloud infrastructure and credentials, involving direct command execution to perform administrative tasks.

Aws Secrets Rotation