Latest AI JobsBrowse Jobs
Obsidian Security logo

Agent Security Research Engineer – Taiwan

Obsidian SecurityTaipei, TaiwanEngineering (D2010)
Apply on company site
Salary
Not listed
Job type
Full-time
Posted
6 Oct 2026
Status
Open · checked 7d ago

About this role

Agent Security Research Engineer – Taiwan is an engineering role based in Taipei, Taiwan, suited for a specialist focused on researching and building security protections for artificial intelligence agents. Obsidian Security seeks an individual to own security content development from initial concept through deployment.

Role Overview

The Agent Security Research Engineer takes charge of agent security content spanning its entire lifecycle. The professional investigates how AI agents, including coding assistants, desktop agents, workflow automation tools, and MCP-connected tools, can be compromised or misbehave. Responsibilities include designing detection mechanisms or policies, building them within engines and pipelines, testing, shipping, documenting, assisting with customer adoption, and tuning based on real-world outcomes.

Key Responsibilities

Research agent misbehaviors and compromise vectors across coding assistants, desktop agents, workflow automation tools, and MCP-connected tools.
Design and build security detection methods and policies within company engines and pipelines.
Test and ship production-ready security content accompanied by documentation.
Assist customers with adoption and tune security content using real-world performance data.
Distinguish real security risks from theoretical risks and articulate the difference to customers, engineers, product groups, and client-facing teams.

Requirements

Practical understanding of agentic tools, including tool calling, hooks, lifecycle events, MCP servers, transports, skills, plugins, and permission modes.
Solid comprehension of agent-specific attack classes such as direct and indirect prompt injection, tool and description poisoning, confused-deputy issues, excessive-agency problems, secret leakage into context, and malicious or over-permissioned MCP servers and extensions.
Strong proficiency in SQL, preferably with analytical stores like ClickHouse, Databricks, or Snowflake.
Discipline in testing by constructing reproducible positive and negative test cases prior to releasing content.
Competency in scripting languages like Python or Go for building test harnesses, parsing event logs, and executing detections.
Practical knowledge of SaaS and cloud identity protocols including OAuth, PATs, API tokens, and scopes, alongside developer tools such as Git, GitHub, GitLab, and CI/CD pipelines.

Nice to Have

Published research, assigned CVEs, conference talks, or blog posts covering LLM or agent security.
Prior familiarity with threat detection concepts.
Practical background in endpoint security on macOS, Linux, or Windows.
Experience utilizing data-pipeline tools like dbt or Dagster.
Professional background in SaaS security, CASB, SSPM, or insider-threat products.
Heavy daily usage of tools such as Claude Code, Copilot, or Cursor with familiarity regarding their security models.

Skills and Tools

Python
Go
SQL
ClickHouse
Databricks
Snowflake
Git
GitHub
GitLab
CI/CD
dbt
Dagster
OAuth

About the Company

Obsidian Security provides cyber security protection for the SaaS and non-human identity layer utilized by modern enterprises. The platform offers unified visibility into human and machine identities, applications, and integrations across SaaS estates, detecting identity-based and supply chain attacks in real time while enforcing least-privilege access. The company serves enterprise clients including T-Mobile, Workday, Snowflake, and S&P Global, and was recognized among Deloitte Technology Fast 500 winners and Forbes America's Best Startup Employers.

How to Apply

Interested candidates should apply directly on the official Obsidian Security careers page.

Skills and tech stack

PythonGoSQLClickHouseDatabricksSnowflakeGitGitHubGitLabdbtDagsterOAuthLLM

Questions candidates ask

What is the location of this job?

This position is located in Taipei, Taiwan.

What kind of systems does the role focus on securing?

The role focuses on agent security content, addressing AI agents such as coding assistants, desktop agents, workflow automation tools, and MCP-connected tools.

What programming languages are required?

Proficient scripting in Python or Go is required to build test harnesses, parse event logs, and perform detections.

Is database or analytical store experience needed?

Yes, strong SQL skills are required, ideally on analytical stores like ClickHouse, Databricks, or Snowflake.

How should an interested candidate apply?

Candidates must apply on the official Obsidian Security careers page.

Ready to apply?

The application is on the official Obsidian Security careers page.

Apply on company site