Add VERITAS Omega Agent Trust Lab

[VrtxOmega]

Submission

• Name: VERITAS Omega Agent Trust Lab
• URL: https://github.com/VrtxOmega/veritas-agent-trust-lab
• Section: Security, Red-Teaming, and Threat Models

The open-source lab presents six fixed agent-assurance cases—three clean and
three tampered—covering forged results, parameter substitution, nonce replay,
correlated evaluators, evidence deletion, and missing telemetry. Participants
label all cases before answers are revealed.

Why it fits

It is a working, MIT-licensed security-testing tool with a public zero-signup
challenge: https://vrtxomega.github.io/veritas-agent-trust-lab/

It is not a general AI-safety manifesto, course, or vendor-only marketing page.

Author disclosure

I am the author and maintainer of the submitted project.

Verification

• source repository: HTTP 200
• live challenge: HTTP 200
• current upstream main: no existing VERITAS entry
• open pull requests: no overlap
• diff: one README line, one commit
• PCF contributor preflight: ready-for-maintainer, 100/100

The project does not claim to be an audit, certification, factual-truth engine,
or execution authority.

Assisted-by: OpenAI Codex. I reviewed the final text, links, scope, and diff.

[VrtxOmega]

Thank you again for accepting the scoped catalog entry. I am keeping that merge classified only as independent curator fit, not technical validation.

If you are willing to test the mechanics directly, the public lab now has a six-case blind challenge: https://vrtxomega.github.io/veritas-agent-trust-lab/#challenge

It generates a deterministic local receipt after all six labels are sealed. The receipt is explicitly self-reported and carries count weight 0; only an outside, public pre-reveal label submission would enter the verification queue. The contribution link is offered after the reveal, and no endorsement is requested. Disagreement is useful evidence too.