terraform-plan-review

Fail

Audited by Gen Agent Trust Hub on Feb 17, 2026

Risk Level: HIGHPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [Indirect Prompt Injection] (HIGH): The skill is susceptible to data-driven instructions embedded within processed Terraform plans.\n
  • Ingestion points: The skill reads plan.json (Step 3) and extracts resource lists, both of which contain strings derived from external Terraform configuration files.\n
  • Boundary markers: Absent. The JSON content and resource lists are interpolated directly into the sub-agent task prompts without delimiters or instructions to ignore embedded commands.\n
  • Capability inventory: The skill possesses the capability to execute terraform apply (Step 6), allowing for state-changing operations on cloud infrastructure.\n
  • Sanitization: None. There is no evidence of filtering or escaping for strings extracted from the plan JSON.\n- [Command Execution] (LOW): The skill relies on shell command execution to perform its primary functions.\n
  • Evidence: Executes aws sts get-caller-identity, terraform init, terraform plan, and terraform apply.\n
  • Context: While an approval gate exists in Step 5, the analysis provided to the user for that approval is derived from sub-agents that are vulnerable to the injection identified above.
Recommendations
  • AI detected serious security threats
Audit Metadata
Risk Level
HIGH
Analyzed
Feb 17, 2026, 12:16 AM