Portfolio
LRPROOFLOOP // AGENT REVIEW LAB

ZERO-TOKEN DEMO

INTERACTIVE WORK SAMPLE / EVALUATION-DRIVEN DELIVERY

What happens after the agent writes the code?

Choose a sanitized software failure, run it through independent reviewer roles, and inspect the fixes, regression tests, and human acceptance evidence.

01

INPUT

Select a failure scenario

CRITICALDOMAIN CORRECTNESS

Payment response misclassification

A gateway returns HTTP 200 for a declined payment. The first implementation treats transport success as payment success and advances the transaction state.

Acceptance criterion
Only a confirmed provider approval may transition a payment to completed.
First-pass gap
No mapping between provider result codes and the domain state machine.
02

CONTROL LOOP

Run independent review

  1. 1BriefGoal + constraints
  2. 2ReviewIndependent roles
  3. 3CorrectFindings → changes
  4. 4TestRegression evidence
  5. 5VerifyHuman acceptance
evaluation.log

READY Select a scenario, then run the review loop.

No external model or paid API is used in this demonstration.

WAITING0 / 5 gates complete

03

JUDGE PANEL

Reviewer verdicts

DOM

Domain reviewer

State and business rules
PENDING

Waiting for an evaluation run.

SEC

Security reviewer

Trust and isolation
PENDING

Waiting for an evaluation run.

UX

Product reviewer

User-visible behavior
PENDING

Waiting for an evaluation run.

TST

Test reviewer

Coverage and regression
PENDING

Waiting for an evaluation run.

04

ACCEPTANCE EVIDENCE

What earns the green light

VERIFIED FIX

Run the loop to reveal the correction.

The accepted solution will appear here with the reason it closes the reviewer finding.

REGRESSION TESTS

  • Evidence is generated after the correction stage.

HUMAN CHECK

  • The final gate remains a human decision.

WHY THIS IS A REAL WORK SAMPLE

Reliable AI engineering is a system, not a prompt.

This lab is intentionally deterministic: the scenarios and verdicts are pre-authored from sanitized engineering patterns, so it works without an API key or token budget.

What it demonstrates is the operating method—explicit acceptance criteria, independent reviewer roles, findings converted into fixes and regression tests, and human accountability at the end.

See the full engineering method