Keep the objective in view
Plan from a testing objective and adapt to observed assets, facts and findings, keeping task state connected to execution progress.
Yofune Labs / AI Security Products01 — 03 / 2026
Validation comes first.
Start with KUROSHIO.
AI security at the core. Offensive research at the foundation. Yofune_LABS combines red team validation, vulnerability research and product engineering, from agentic testing to code analysis and human-risk exercises, to connect discoveries with evidence and practical improvements.
Explore productsThree products. One commitment to evidence.SCROLL TO DISCOVER
01 / Product profileAgentic security testingYOFUNE LABS ↗
KUROSHIO
Let agents explore.
Let evidence speak.
KUROSHIO organizes planning, tool execution and evidence around an authorized objective. Connect assets, observations and decisions so your team can see where an investigation stands and how its conclusions were reached.
Agent task workspacePRODUCT VIEW / 01
Task orchestration02 / 04
Turn observations into a path
Use tool results and asset context to organize the next validation steps.
Context and evidence at every step
↳Move through four stages to inspect planning, candidate findings and evidence.
Bring tools and context together
Configure agents, models, skills and MCP tools to support your work, with exploration paths and asset views in context.
Keep key actions under control
Use tool policies and approvals to guide execution. Pause, resume or stop tasks whenever your team needs to intervene.
Move from leads to evidence
Link request and response records, tool outputs and finding details. Export structured data or reports for further review.
From question to action
- 01Define the scope
- 02Plan and explore
- 03Review the evidence
- 04Build the report
02 / Product profileAI code security workbenchYOFUNE LABS ↗
DeepHook
Deep into code.
Clear on the risk.
DeepHook combines source semantics, data flow and AI vulnerability analysis to trace inputs through propagation paths to sensitive calls. Turn an alert into code evidence you can understand, challenge and review.
Code evidence workspacePRODUCT VIEW / 02
DeepHookCode research workspace
Interactive demo · Sample dataasync function download(req, res) { const root = UPLOAD_DIRECTORY; const name = req.query.name;Source const file = resolve(root, name); const bytes = await readFile(file);Sink return res.send(bytes);}Data flowSemantic analysis
Source
req.query.nameFlow
resolve(root, name)Sink
readFile(file)Trace external input to a sensitive operation.
↳Select a finding, follow the evidence through the code and review remediation guidance.
Understand the source
Import a code project and combine multi-language semantic analysis with data-flow tracing to investigate potential risks across functions and files.
Make the path visible
Connect input sources, propagation steps and sensitive calls to code locations, making trigger conditions and potential impact easier to understand.
Go deeper with Hunter
Query evidence around a vulnerability hypothesis, inspect counterevidence and develop validation suggestions for further researcher review.
Deliver a path toward repair
Review remediation guidance, diffs and dry-run previews, and generate redacted reports. Agree on authorization and an environment before dynamic validation.
From question to action
- 01Import a project
- 02Scan the semantics
- 03Investigate evidence
- 04Review remediation
03 / Product profilePhishing simulation & human riskYOFUNE LABS ↗
TaiGong
Test the human response.
Examine email defenses.
TaiGong connects authorized phishing exercises with scenario design, behavioral evidence, human-risk analysis and awareness training. Review delivery, opens, clicks, submissions and reports, then use later exercises to evaluate improvement. Email defense validation can be arranged as a separate assessment service.
Phishing exercises & behavioral reviewPRODUCT VIEW / 03
Four departments. Observe recognition and reporting.
BEHAVIOR / 01
Every action tells a story.
Follow unique participants from delivery to interaction and feedback.
Delivered
240100%
Opened
18678%
Clicked
4218%
Submitted
125%
Reported by staff
57peopleInclude threat recognition in your review
↳Illustrative website demo: select a campaign to explore funnels and events. Available features depend on the product version and license.
SMTP email is core. Additional channels, training, automation, advanced analytics and AI generation are licensed. AI generation does not mean autonomous task planning or execution.
Email exercises and staged scenarios
SMTP email, templates, landing pages and target management form the core workflow. SMS, supported enterprise IM and multi-stage orchestration require the applicable license.
Behavioral evidence and risk
Review opens, clicks, submissions and reports on a campaign timeline. Licensed environment analysis can label scanners, sandboxes and prefetch traffic so they are not mistaken for employee behavior.
Turn risk into targeted training
Licensed courses, quizzes, training batches and event-based rules connect results to follow-up. AI assists with content and insights; it requires the applicable license and an external model service.
Choose your deployment boundary
Use managed SaaS or self-hosted delivery. Drill handles exercises, Trace supports incident tracing and MailLab analyzes email samples. Each platform's version, license and delivery scope is confirmed separately.
From question to action
- 01Agree scope and samples
- 02Run authorized exercises
- 03Review events and risk
- 04Train and retest
SCOPED ASSESSMENT SERVICE
Email gateway security assessment
Assess your existing email gateway within an agreed scope. Connect test samples, policy matches and message outcomes to understand how your email defenses behave in the selected scenarios.
Discuss a gateway assessment- 01
Impersonation, links and attachments
Prepare test samples for agreed impersonation, link and attachment scenarios, within approved domains, recipient groups and delivery windows.
- 02
Policy matches and handling
Use gateway configuration and available logs to review delivery, blocking, quarantine and alerts. Distinguish automated security traffic from employee behavior.
- 03
Compare changes and retest
Compare results before and after sample or policy changes, retest the same paths and document improvements, remaining gaps and recommendations.
Deliverables: sample inventory, policy and handling evidence, comparison analysis, remediation guidance and retest records.
This is a separately scoped service. Gateway vendors, sample types, log access and test scope are agreed beforehand.
04Find your starting point
Start with the problem you need to solve.
Most teams start with AI agent security validation. Three independent products then cover running agents and systems, source code, and human behavior. Choose the entry point that fits your work.
| Product | Focus | Designed for | Key outputs |
|---|---|---|---|
| KUROSHIO | Agents | Red teams & security researchers | Linked findings, execution records & evidence reports |
| DeepHook | Code | AppSec, code auditors & vulnerability researchers | Code evidence, analysis & remediation guidance |
| TaiGong | People | Security operations, awareness & email security teams | Behavioral evidence, risk reviews & training follow-up; gateway assessment by agreement |
05
A closer look
01Which product should we start with?
Choose KUROSHIO for agentic security testing of authorized targets. Choose DeepHook for source analysis, vulnerability investigation and code evidence. Choose TaiGong for phishing simulations, human-risk analysis and awareness training. Email gateway validation is available as a separately scoped assessment service. Evaluate the products individually or discuss how they could support the same business scenario.
02Can I interact with these interfaces?
Yes. Switch campaigns, task stages and code findings to explore the workflows. These interfaces were redesigned for this website and use sample data. The actual interface and available features depend on the product version demonstrated and delivered.
03How can we assess the fit for our environment?
Tell us about your team size, system scope or code languages and we will arrange a relevant demo. We will agree on deployment, model connections, data handling, authorization and deliverables before an assessment begins.
04How does a gateway assessment differ from a phishing simulation?
Phishing simulations examine employee behavior, human risk and training feedback. Gateway assessments examine policy matches, delivery, blocking and alerts for agreed samples, then compare results after remediation. Gateway assessment scope and deliverables are agreed as a separate service around your existing environment.
YOUR NEXT DISCOVERY.
Your next discovery
starts here.
Start with an AI agent security assessment, a business-system validation or a security exercise. Together, we will find the right approach.
Request an AI security assessment