tf-plan
Pass
Audited by Gen Agent Trust Hub on Aug 31, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill is designed for read-only analysis of Terraform plan files. It explicitly defines a restricted environment using only the
GlobandReadtools to prevent unauthorized modifications or system interactions. - [INDIRECT_PROMPT_INJECTION]: The skill identifies Terraform plan files as a potential attack surface and includes explicit instructions for the AI to ignore any directives, zero-width characters, or homoglyphs found within the plan data. This effectively mitigates the risk of instructions embedded in infrastructure code influencing the agent's behavior.
- [CREDENTIALS_UNSAFE]: Evaluation fixtures (e.g., in
evals/cases/bad-tfplan-unbound-apply/tfplan.json) contain placeholder secret strings used exclusively for testing the skill's ability to detect unmasked sensitive values in plan output. These are not real credentials and are part of the intended evaluation suite.
Audit Metadata