Security Reviews for High-Consequence Systems
Protocols · Agentic Systems · Incident Reconstruction
I run an independent, founder-led security research and engineering practice. I work with protocol teams and builders of tool-using AI systems where failures can cross code, permissions, verifiers, external data, and on-chain execution.
My work is evidence-led: define the scope, identify the trust and authority boundaries, test concrete failure paths, preserve reproducible evidence, and state what remains unresolved.
| Area | Focus | Typical output |
|---|---|---|
| Protocol Security Reviews | Smart contracts, protocol invariants, authorization, accounting, callbacks, upgrades, and verifier/application boundaries | Threat model, review ledger, reproducible findings, technical report, remediation review |
| Agentic Systems Security Reviews | Tool permissions, untrusted context, indirect prompt injection, data exposure, goal hijacking, authority escalation, and transaction policy | Authority map, adversarial cases, evidence report, prioritized hardening and regression plan |
| Incident Reconstruction | On-chain execution, asset movement, contract relationships, root-cause analysis, provenance, and evidence gaps | Bounded chronology, evidence bundle, technical report, limitations and unresolved questions |
Engagement scope, revision, exclusions, confidentiality, delivery dates, and commercial terms are agreed before work begins. Review the engagement model →
| Work | Evidence |
|---|---|
| Security Reviews | Three published protocol reviews with explicit scope, findings, and limitations |
| Security Review Skills | Versioned, evidence-gated workflows for EVM, proof-verifier, and zero-knowledge review |
| Agent Authority Lab | Reproducible recipient-substitution and approval-binding case for transaction authority |
| White Radar | Read-only EVM incident reconstruction with bounded evidence collection and provenance |
| ZKBind | Cross-layer analysis of verifier-to-application trust boundaries |
| Web3 Security Corpus Builder | Reproducible collection, deduplication, FTS5 indexing, and RAG export pipeline |
Evidence over assertion · Reproducibility over narrative · Explicit limits over false certainty
All work is defensive and authorized. A security review reduces uncertainty within a defined scope; it does not guarantee the absence of vulnerabilities or future incidents.