The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: Automated scanners have blacklisted the documentation URL hosted on the vendor's site: https://jeffallan.github.io/claude-skills/skills/infrastructure/terraform-engineer/. The skill also recommends downloading third-party tools like tflint and pre-commit hooks.
[CREDENTIALS_UNSAFE]: The references/providers.md file explicitly references the standard AWS credentials path ~/.aws/credentials as a configuration option, creating a surface for sensitive file access.
[COMMAND_EXECUTION]: The workflow requires the agent to execute shell commands like terraform plan and terraform apply. This capability allows for significant infrastructure changes and potential system access if given malicious code.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process Terraform configuration files and external documentation, which represents an attack surface for indirect prompt injection.
Ingestion points: Processes Terraform code files and local documentation references in markdown format.
Boundary markers: Implements a manual approval step where the agent must present a plan and wait for user confirmation before executing terraform apply.
Capability inventory: Extensive system access through shell execution of terraform and related ecosystem tools.
Sanitization: Relies on standard CLI validation tools (terraform validate, tflint) to check for syntax and linting errors rather than security-focused sanitization of inputs.

terraform-engineer