The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The script programmatically accesses and verifies AWS credentials using get_aws_client() and client.validate_credentials(). This grants the skill direct access to the user's cloud identity and permissions.
[DATA_EXFILTRATION]: The skill executes discover_sbom(client), which scans the user's AWS environment to build a detailed tech stack profile. This sensitive infrastructure information is then passed to generate_daily_advisory, which typically involves external network operations to query threat intelligence feeds, potentially exposing the environment's configuration to external services.
[COMMAND_EXECUTION]: The skill executes a multi-line Python script using the -c flag. It dynamically determines the command to run by reading python_cmd from shasta.config.json, which could lead to arbitrary command injection if the configuration file is controlled by an untrusted source.

threat-advisory