Skip to content
Current page: Technical Architecture Audit Agent Workflow

Run an evidence-backed technical architecture audit

Bound an architecture question, gather repository evidence in read-only mode, compare options, and review the recommendation before implementation.

A technical audit is most useful when it resolves a defined decision and leaves behind inspectable evidence, trade-offs, and open questions—not an unbounded opinion about the codebase.

Set the audit boundary

Name the architecture, performance, reliability, or security question; components and environments in scope; explicit exclusions; evidence to inspect; and decision criteria for comparing options.

Use a read-only research pass

The technical-audit starter recipe assigns Research to inspect the system and produce findings, impact, options, trade-offs, and a recommendation without changing files. A Reviewer then challenges its assumptions and evidence.

Separate decision from implementation

The team accepts, rejects, or revises the recommendation before opening implementation work. This preserves a clear record of why an option was chosen and prevents an agent from turning an unreviewed audit opinion directly into code.

Bring every coding agent onto one shared plane.

Plan with your team, run Claude and Codex in isolated sandboxes, and review the work before it ships.

Start a project