SlipstreamJobs tracks this role from the company's public career site. Apply directly on the employer's site.
Runlayer is building the enterprise platform for AI agents, MCPs (Model Context Protocol), and skills with built-in security, governance, and observability. The company recently raised $30M Series A (Felicis, Khosla Ventures) and is trusted by Gusto, Instacart, Opendoor, dbt Labs, and Decagon.
As the first Security Researcher, you will identify and exploit vulnerabilities across the AI agent ecosystem—MCP servers and clients, AI coding agents, agent frameworks, skills and plugin marketplaces, and OAuth flows connecting them. You will own the full research lifecycle: vulnerability discovery, coordinated disclosure with vendors, CVE/advisory publication, and responsible embargoes. Your findings will be published as technical write-ups, open-source tools, and conference talks that shape industry thinking on agent security. You will also translate research into product protections: detection rules, scanner capabilities, and public risk ratings for MCP servers.
You will run ecosystem-scale studies across thousands of MCP servers using Runlayer's catalog and scanning infrastructure. You will brief customers, prospects, and press alongside marketing and developer relations teams. You will contribute to MCP specification security work and help establish industry frameworks and best practices.
This role offers significant impact: your research will directly influence how enterprises, vendors, and standards bodies approach agent security, and every finding ships as a protection for Runlayer's customers.
REQUIREMENTS:
- 5+ years in offensive security research, vulnerability research, or red teaming
- Public record of security work: published CVEs or advisories, conference talks, or open-source security tools and write-ups
- Deep expertise in agent-native attacks: indirect prompt injection through tool output, tool poisoning, cross-server shadowing, confused deputies via OAuth, supply-chain attacks on skills and plugins
- Proficiency writing Python, TypeScript, or Go for harnesses, fuzzers, and scanners
- Clear technical writing for both engineers and security leaders
- Sound judgment on responsible disclosure, including negotiating with resistant vendors
- Daily hands-on use of AI agents as both tools and attack targets
BONUS QUALIFICATIONS:
- Published research on LLM, agent, or MCP security
- Prior experience on a security vendor's research team or at an offensive security consultancy
- Speaking engagements at Black Hat, DEF CON, RSA, or equivalent conferences
- Established relationships with vendor security response teams and security media
- Open-source security tools with active user bases
