The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes infrastructure configuration data from local HCL files which influence the execution of high-privilege infrastructure commands.
Ingestion points: Host and network configuration files (inventory.hcl, networking.hcl, versions.hcl) are read via read_terragrunt_config to populate module inputs.
Boundary markers: The instructions explicitly require human approval for all task tg:apply operations and mandate manual verification of hardware addresses and IP allocations.
Capability inventory: The skill enables infrastructure provisioning and state modification using terragrunt and tofu through task command wrappers.
Sanitization: Relies on manual human review of generated plans and HCL syntax validation before execution.
[EXTERNAL_DOWNLOADS]: The skill fetches infrastructure modules and stack definitions from external repositories.
Evidence: Examples in references/stacks.md and references/units.md utilize git::github.com sources for Terragrunt stacks and modules. This represents standard infrastructure-as-code dependency management using well-known services.
[COMMAND_EXECUTION]: The skill instructs the agent to execute shell commands via a task runner to manage infrastructure state and deployment.
Evidence: SKILL.md defines several task tg:<command> aliases for validation and deployment. It also provides specific safe patterns for resource state management using bash loops to preserve resource identifiers.

terragrunt