Reading a Bluebricks PR

Understand the structured PR format that the Bluebricks agent uses when opening pull requests for infrastructure changes

Overview

When the Bluebricks agent makes infrastructure changes, it opens a pull request with a structured description designed for fast, confident reviews.

New to Bluebricks? Bluebricks is an agentic infrastructure platform. The agent can discover cloud resources, generate infrastructure code, and open pull requests on your behalf. Learn more in What is Bluebricks?

How the agent opens PRs

A typical flow looks like this:

1. You ask the agent to make an infrastructure change (for example, codifying unmanaged cloud resources or modifying an existing configuration)

2. The agent generates or updates infrastructure code, validates it against the live environment, and iterates until the plan is stable

3. The agent opens a PR with a structured description so reviewers can evaluate the change quickly

The PR description is built around three questions a reviewer needs to answer:

• Did the agent do what I asked?

• Did it touch only what it should?

• Is it safe to merge?

Each section of the PR maps to one or more of these questions.

PR sections

Every agent-generated PR includes some or all of the following sections. The agent only includes sections where it has concrete data; empty sections are omitted entirely.

Summary: what the agent changed

A one-line description of what the agent changed, written in past tense. For complex changes that span multiple resources, the summary includes bullets listing each modification.

Answers: did the agent do what I asked?

Blast radius: risk assessment and scope of impact

A risk assessment in exactly three lines:

• Nature of change (first word): describes what kind of change the diff contains

Label	Meaning
Additive	New resources only; nothing existing was modified or removed
Modification	Existing resources updated in place
Destructive	Resources removed or recreated

• Downtime risk: none, low, medium, or high, with an explanation of why

• Confidence: how well the agent understands the change and its impact

Level	Meaning
High	The agent fully understands this change and its blast radius
Medium	The agent understands the change but not all side effects
Low	The agent is not confident about the impact

Every label includes a justification. The agent is required to explain its reasoning so reviewers can audit the assessment.

Answers: did it touch only what it should? and is it safe to merge?

Materials: resources and files involved

A reference table listing the resources and files involved in the change:

Field	Description
Terraform address	The resource address in code (e.g., aws_s3_bucket.main)
File	The file containing the resource definition
Resource ID	The cloud resource identifier
Region	Where the resource is deployed
Provider	The cloud provider (aws, azure, gcp)

For changes that span multiple files, a "Files changed" line appears below the table.

Answers: did it touch only what it should?

Inspection: how to verify the change

A set of concrete verification steps the reviewer can run to confirm the change works as expected. Each row in the table includes a check name, the command or action to run, and the expected result.

Answers: is it safe to merge?

References: related external documentation

An optional list of links to relevant external documentation (Terraform provider docs, cloud provider references, or Bluebricks docs). This section only appears when the agent can provide valid URLs; it is omitted rather than showing placeholders.

Next steps

• Learn more about the platform: What is Bluebricks?

• Learn how the agent works: Agent Overview

• See how the agent codifies cloud resources: Codifying Infrastructure