The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to process and interpret external data sources, specifically Terraform plan files and CI annotations, which represents a potential injection surface.
Ingestion points: Terraform plan JSON files processed by the Python script in gpt.md, and plan screenshots/CI output interpreted in gemini.md.
Boundary markers: The skill includes mandatory directives for human review before applying changes and specific warnings about blast radius classification.
Capability inventory: The skill uses the terraform CLI to perform infrastructure changes (terraform apply in SKILL.md).
Sanitization: The provided Python script implements a safety gate to detect and block 'delete' actions on protected resource types.
[EXTERNAL_DOWNLOADS]: The skill references external resources and community tools for infrastructure management.
Fetches documentation and best practices from HashiCorp's official developer portal.
References community-maintained resources such as terraform-gotchas and the tfautomv utility on GitHub for refactoring guidance.

terraform