The Agent Skills Directory

[DATA_EXPOSURE]: The documentation in references/providers.md refers to sensitive file paths such as ~/.aws/credentials and service-account-key.json for provider authentication. While these are standard Terraform authentication methods, they represent sensitive local data that an agent would interact with.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to process user-provided Terraform configurations and modules which could contain malicious instructions or local-exec provisioners.
Ingestion points: Terraform configuration files (.tf), variable files (.tfvars), and external module sources defined in references/module-patterns.md.
Boundary markers: None explicitly defined to separate instructions from code-driven data.
Capability inventory: Execution of terraform (including plan/apply), tflint, go test (which compiles and executes test code), and package managers like pip and brew.
Sanitization: Relies on terraform validate and terraform fmt to ensure HCL syntax compliance, but these do not sanitize against logic-based prompt injection.
[EXTERNAL_DOWNLOADS]: The skill provides instructions for installing external tools and dependencies in references/testing.md using brew install tflint, pip install pre-commit, and go mod download. It also references remote Git modules in references/module-patterns.md using git::https://github.com/org/terraform-modules.git.

terraform-engineer