Skip to content
Current page: Technical Audit and Recommendation AI Recipe

Audit the system before choosing a technical change

Run a read-only architecture, performance, reliability, or security audit and challenge its evidence before code changes begin.

This starter recipe is for technical questions that should begin with evidence and a recommendation, not an implementation.

The agent chain

Research performs a read-only audit at ultracode effort and documents findings, impact, options, trade-offs, and a recommended next step. Reviewer follows at high effort to challenge the evidence, assumptions, prioritization, and recommendation.

Bound the audit

State the decision or risk the audit must resolve, the components and environments in scope, explicit exclusions, the code paths or measurements to inspect, and the criteria for comparing options.

  • Architecture boundaries and dependencies
  • Performance, reliability, or security evidence
  • Trade-offs and constraints for each viable option
  • Open questions that still require a human decision

The output is a recommendation

This recipe intentionally does not begin with code changes. The team reviews the recommendation and can create a separate implementation card or recipe after deciding which option to pursue.

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