The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data from Terraform configuration files (.tf, .tfvars), JSON variable files, and flow definitions. Ingestion points: Identified in SKILL.md as file contents read from disk. Boundary markers: The skill contains explicit instructions to treat these files as data rather than instructions and to flag any embedded directives like 'ignore previous instructions'. Capability inventory: The skill restricts the agent from executing commands, only allowing it to provide text for manual user execution. Sanitization: The instructions advocate for strict type-checking and format validation of all input values.
[COMMAND_EXECUTION]: The skill implements a defensive constraint by instructing the agent never to execute Terraform commands directly. Instead, it must provide the exact shell commands for the user to review and run, ensuring human-in-the-loop oversight for infrastructure changes.
[SAFE]: The skill follows security best practices for secret management by recommending Terraform variables for management keys and avoiding hardcoded credentials. It utilizes the official Descope Terraform provider from the official Terraform registry.

descope-terraform