The Agent Skills Directory

[PROMPT_INJECTION]: The skill processes untrusted Terraform HCL code, which presents an attack surface for indirect prompt injection if the code contains malicious instructions or configurations.
Ingestion points: User-provided infrastructure-as-code files and project structures evaluated by the agent.
Boundary markers: Instructions do not define specific delimiters or provide "ignore embedded instructions" directives for the HCL content being processed.
Capability inventory: The skill leverages the terraform CLI, which can execute arbitrary shell commands via local-exec provisioners or external data sources.
Sanitization: No explicit sanitization or validation of the input HCL content is described.
[COMMAND_EXECUTION]: The skill instructs the agent to automatically execute a shell command (ls) upon activation to check for the presence of recommended companion skills in the user's environment.
[DATA_EXFILTRATION]: The activation routine performs reconnaissance on the user's filesystem by checking for the existence of directories in the home folder (e.g., ~/.claude/skills/), which exposes metadata about the user's local configuration.

terraform-iac