Repeatability
Medium
The structural pattern — parse state, find untagged resources, emit remediation HCL — is consistent. But each organization's tagging taxonomy, resource types, and state file layout differ enough that the script needs meaningful customization per engagement.
Ambiguity Tolerance
Low
Success criteria are underspecified: what tags are required, what values are acceptable, which environments are in scope, and what 'standardized' means are all undefined. The agent must either guess or ask, and guessing wrong on a prod remediation file is costly.
Data & Tool Availability
Low
The agent has no access to the actual Terraform state files, AWS credentials, or the 8 environment configurations — these must be provided explicitly. Without them, the agent can only produce a generic template, not a working audit against real infrastructure.
Error Cost
High
A malformed or incorrect remediation Terraform file applied to 200+ resources across production environments could trigger unintended resource modifications, state corruption, or outages. Even tag-only changes can cause policy violations or cost allocation breakage.
Human Judgment Required
Medium
Defining the tagging taxonomy (cost center, owner, environment values) requires organizational knowledge the agent lacks. However, once the schema is provided, the mechanical work of identifying gaps and generating HCL is well within AI capability.