Phantom ↗
A browser extension for frontend asset discovery and API mapping during authorized security testing.
Recursive JavaScript scanning, configurable concurrency, batch API checks, custom rules, and structured exports.
I research application and AI agent security, and build tools for reproducible security work.
Java / Spring code review, Agent / MCP trust boundaries, and coordinated vulnerability disclosure.
Security tools and agent workflows, developed independently and with SoybeanSeclab. Repository stars as of November 14, 2026.
A browser extension for frontend asset discovery and API mapping during authorized security testing.
Recursive JavaScript scanning, configurable concurrency, batch API checks, custom rules, and structured exports.
A local Python CLI that organizes the work around AI coding agents, from task context to review.
Builds context packs and agent instruction files; records commands, logs, diffs, checks, and review reports. Current release: v0.1.0, CLI-first.
A four-role CTF agent system with asynchronous collaboration, shared budgets, and session artifacts.
Recent work improves evidence tracking, per-round records, interruption handling, HTTP observation, and experiment isolation.
A lightweight Go command-and-control framework for authorized security labs and exercises.
Combines a web console, peer-node coordination, task and result management, file workflows, and Docker deployment.
A Java / Spring audit workbench connecting route and permission context with source review.
v1.2.0 supports Semgrep / CodeQL SARIF import, heuristic checks, a local GUI, AI-assisted review, and JSON / Markdown / HTML / SARIF reports.
Selected public findings and coordinated reporting. Advisory status checked November 14, 2026.
Server-side request forgery via integration remote fetch. The official advisory credits OracleNep as the reporter.
The advisory documents internal-network access and integration-secret exposure, with remediation across releases 2026.08.0 and 2026.10.1.
Research into authentication and trust boundaries in the dsh web agent control plane, tested on 0.1.0-rc.6.
The report distinguishes local-process access from remote exposure when the service is configured to listen beyond loopback, and proposes authentication and authorization improvements.
Submitted vulnerability reports through GitHub's private security advisory workflow for Flowise and Strapi.
These reports are not published advisories. Technical details remain private during coordination with maintainers.
Selected identifiers from individual and team vulnerability reporting.
Selected patches with public review and merge history.
Added bounded recovery when the solve engine exhausts its open intents before completing the goal, with fallback exploration, CLI events, and regression coverage.
Documentation and test-comment maintenance for error conventions in ext/intl.
Aligns request_parse_body() handling of max_input_vars with INI behavior, with PHPT regression coverage and compatibility notes.
From source review and reproducible findings to remediation and retesting.
I review Java / Spring applications by tracing request entry points, permission checks, data flow, and dangerous operations. My tooling work connects source context with Semgrep / CodeQL findings and SARIF reports.
My practical experience includes authorized vulnerability research, incident triage, host-security checks, impact assessment, and remediation follow-up.
I study where agents cross trust boundaries: tool execution, MCP permissions, local control planes, and approval flows. I also work on recovery behavior, evidence records, and human review in agent-assisted security workflows.
Research notes document prerequisites and verification limits alongside findings.