About
What is tier-zero?
The product
tier-zero is an open-source multi-agent system that audits a GitHub profile and produces an opinion in the voice of a staff engineer. It is built for engineers who want an honest mirror of how their public work reads to a senior reviewer.
How it works
Five specialised agents run in a graph: Forensics (originality and authenticity), Depth (technical depth and production markers), Claims (bio versus code alignment), Senior Reviewer (verdict synthesis), and Critic (adversarial review). The Critic challenges the Reviewer until the verdict survives scrutiny, capped at two revision rounds.
What it isn't
tier-zero is not a hiring decision, a credit score, or a factual statement about any person. Every output is an opinion produced by a large language model and must not be relied on as ground truth. tier-zero may be wrong.
Self-audit only
v1 only audits the GitHub profile of the user requesting the audit. Auditing other people's profiles is restricted by design. We open up other modes only when we are confident the safety guarantees hold.
Open source
Methodology, prompts, eval set, and source code are all public at github.com/KrishanKVerma/tier-zero. You can read exactly how a verdict is produced.
Author
tier-zero is built by Krishan Kumar Verma, a final-year B.Tech student. The project is in active development and the verdicts are not always sharp. Feedback welcome.