The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The skill executes gcloud iam service-accounts keys create ./gcp-credentials-local.json. This command generates a persistent, sensitive private key file on the local filesystem. While the skill warns the user and suggests adding it to .gitignore, the automated creation of long-lived, high-privilege credentials on disk is a high-risk security practice.
[COMMAND_EXECUTION]: The skill relies extensively on shell command execution via the gcloud CLI to enable APIs, create repositories, manage service accounts, and modify IAM policies. These operations have a high impact on the security posture of the target GCP environment.
[EXTERNAL_DOWNLOADS]: The skill provides links to official installation sources for the Google Cloud SDK and Homebrew. These are well-known technology services and are documented here for transparency.
[PROMPT_INJECTION]: Indirect prompt injection surface (Category 8). The skill reads untrusted data from a local configuration file and uses it in shell commands without evidence of sanitization or strict validation.
Ingestion points: .spring-config.json (Step 0)
Boundary markers: None implemented for configuration variables.
Capability inventory: Extensive subprocess execution via gcloud (Steps 3-7).
Sanitization: No evidence of shell-escaping or input validation for variables like app_name or gcp_project_id before interpolation into shell commands.

springboot-gcp-setup