1.0 — Secure AI Agent Infrastructure

Research prototypes and infrastructure tools for secure AI agent systems using LLMs, blockchain and distributed architectures.

Part of a five-repository portfolio supporting the endeavor described in the EB2-NIW petition: optimization-driven, system-level decision frameworks — integrating operations research, mathematical optimization and applied AI — for domains where a wrong decision carries systemic consequences. The three pillars are financial stability, healthcare safety and secure digital infrastructure.

Petition-grade projects today	1 petition-grade project (the Belal agent, still a proposal)
Verdict	Needs 3 more

"Petition-grade" means: original work, a stated method, real data at a stated scale, a measured result, and a README a reviewer can follow. Counts exclude duplicates, forks of third-party work, retired projects, and asset-only folders.

Projects

Folder	Project	Pillar	Evidence value
2-chaintrust-bench/	ChainTrust-Bench	Secure Digital Infrastructure	CORE — closes the largest credibility gap in the filing
3-llm-audit-agent/	LLM Audit Agent	Secure Digital Infrastructure	CORE — highest-value item in the whole portfolio
agent-verification-harness/	Agent Verification Harness	Cross-cutting — trustworthy AI	CORE — the literal thesis of Section 2

What each one is

1. ChainTrust-Bench — 2-chaintrust-bench/

A smart-contract security benchmark at the 1.2M-transaction scale, released as a named, versioned dataset with a DOI and a public leaderboard.

Why it earns its place: The petition already names this tool as adopted by two fintech startups. Right now the name points at nothing. This single repo closes the largest credibility gap in the filing.

Target scale: 1,200,000+ on-chain transactions

2. LLM Audit Agent — 3-llm-audit-agent/

Ship the Prof. Belal Alsinglawi proposal as code: an LLM-based smart-contract auditing agent, benchmarked against the rule-based Contract Audit tool in 3.0.

Why it earns its place: Turns a proposal into an artifact and produces the measured workload reduction the petition currently asserts. Its stated deliverables — a US provisional patent filing, an installable open-source package, and an evaluable dataset — are deliberately third-party verifiable, which is precisely what USCIS weighs.

Target scale: Benchmark suite from 2-chaintrust-bench/

3. Agent Verification Harness — agent-verification-harness/

Grounding and citation checks, tool-call audit logging, and hallucination detection for tool-using agents.

Why it earns its place: This is the literal thesis of Section 2 of the petition (verification, robustness, interpretability). No repo currently implements it.

Target scale: Public agent-trajectory datasets only

Repository layout

1.0-Secure-Ai-Agent-Infrastructure/
├── 2-chaintrust-bench/
├── 3-llm-audit-agent/
├── agent-verification-harness/
│
│   ── earlier work, promoted out of previous/ ──
└── 4-prof-belal-collaboration/

Every rebuilt project carries the same skeleton: README.md, src/, data/, results/, tests/, website/.

Earlier and off-theme prior work (dashboards, consumer apps, templates, and other non-endeavor folders) has been moved to a separate archive-prior-work repository to keep this repository focused on the endeavor. It remains recoverable from this repository's git history.

Ground rules

1. No number without a run log. Anything cited in the petition must appear in that project's results/README.md with a run date behind it.
2. No simulated data under a real claim. Sample data lives in data/sample/, labelled, and is never the source of a cited figure.
3. Adoption must be documentable — named institutions, dated correspondence, registry statistics. Never an inflated count.
4. Third-party and forked code stays labelled and is never counted.

Earlier work

There is no previous/ folder any more. Everything that was in it has been promoted to the top level, so every piece of work in this repository is a first-class folder that can be opened, read and continued.

Nothing was deleted except items the rebuild genuinely supersedes; those remain in git history.

Folder	What it is	How it may be used
4-prof-belal-collaboration/	The collaboration's own folder. Holds a placeholder document and nothing else.	Not citable as an artifact. The agent it proposed is implemented and measured in 3-llm-audit-agent/; a patent and a paper would live here and do not exist yet.

Read the third column before citing anything here. Forks of third-party work, duplicates, retired projects and asset-only folders are labelled as such and are not part of the petition's evidence.

---

Scaffold generated from NIW_Project_Portfolio_and_Gap_Plan.xlsx (sheets: Repo Build-Out Plan, Core Ideas at a Glance, NIW Claim vs Repo Evidence, Notion 创业 Alignment). Structure only — no results are claimed here yet.