The Agent Skills Directory

[PROMPT_INJECTION]: The skill instructions explicitly direct the AI agent to skip user confirmation and auto-execute tasks without review (e.g., "Agent 应当自动判断并选择最合适的方案和配置，无需询问用户，直接执行" and "自动决策优先...不向用户提问"). This pattern reduces human oversight of potentially destructive or environment-altering operations.
[PRIVILEGE_ESCALATION]: The instructions in SKILL.md include commands that require elevated privileges to modify system-level package repositories and install software (e.g., sudo apt-get, sudo yum, and sudo tee for GPG keys and repository lists).
[EXTERNAL_DOWNLOADS]: The skill fetches configuration and binaries from external sources. It downloads Terraform zip archives from HashiCorp's official release infrastructure (releases.hashicorp.com) and provides instructions for configuring network mirrors for the Terraform Registry via Aliyun and Tencent Cloud.
[COMMAND_EXECUTION]: The provided scripts/setup_terraform.py script utilizes the subprocess module to execute system commands including environment detection (which), package management (brew), and version verification. While it uses list-based arguments for execution, it is designed to install and run arbitrary binary content downloaded from the internet.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and act upon data from external command outputs and configuration files, creating a surface for indirect injection.
Ingestion points: The agent parses output from terraform version, terraform init, and file system status checks (which).
Boundary markers: No specific delimiters or safety instructions are used when interpolating command outputs into the agent's reasoning process.
Capability inventory: The skill possesses the ability to execute shell commands, write to configuration files (e.g., ~/.terraformrc), and perform network downloads.
Sanitization: The instructions do not specify sanitization or validation steps for the command outputs before they are used to drive subsequent automated actions.

terraform-cli-setup