Tenable launches Exchange Inspector to vet community AI agents, skills and MCP servers on CyberAgents Exchange
Tenable has detailed the Exchange Inspector, a three-stage vetting process for community-contributed AI agents, skills, MCP servers and playbooks listed on the open-source CyberAgents Exchange. The process was developed through Tenable's participation in the OpenAI Daybreak Defense Network. Three listings have passed so far: two from Tenable (SOC-Hunter and the Remediation Priority & Impact Agent) and one from Splunk (Splunk Tenable Cloud Security Skill). This is product and industry news with no incident, vulnerability or active threat described. It is relevant to defenders assessing supply chain risk when deploying agentic AI in SOC workflows.
Detalhes técnicos
- Stage 1: Automated screening with the skills-inspection engine in Tenable One AI Exposure. It parses agent instructions, authorized tools and data access, and flags prompt injection, jailbreaks, hidden instructions, hardcoded secrets, PII exposure and sensitive data access.
- Stage 2: Frontier assessment with OpenAI GPT Cyber models via Daybreak. Daybreak Blue supports source code review and dual-use components. Daybreak Red is reserved for higher-risk submissions that need exploit validation and adversarial testing.
- Stage 3: Tenable researchers validate findings and run the component in a clean environment to verify runtime behavior. They produce an auditable record covering provenance, threat model, source review and runtime verification.
- The review covers 15 issue classes across the model, application and infrastructure layers. Examples are prompt injection, excessive agency, memory poisoning, secrets handling, SSRF, path traversal, command execution and supply chain risks such as typosquatting and unpinned packages.
- Tenable reports that SOC-Hunter, a Claude Code skill, cut hunt times by 75% (from 4-6 hours to 45-90 minutes).
Remediação
- Prefer vetted listings, and review the public GitHub source of any community agent, skill or MCP server before production use.
- Assess the permissions, credentials and tool access of AI agents already running in your environment, for example with the skills inspection in Tenable One AI Exposure.
Linha do tempo
-
Tenable launches Exchange Inspector to vet community AI agents, skills and MCP servers on CyberAgents Exchange
ler a análise desta fonte ocultar
Tenable has detailed the Exchange Inspector, a three-stage vetting process for community-contributed AI agents, skills, MCP servers and playbooks listed on the open-source CyberAgents Exchange. The process was developed through Tenable's participation in the OpenAI Daybreak Defense Network. Three listings have passed so far: two from Tenable (SOC-Hunter and the Remediation Priority & Impact Agent) and one from Splunk (Splunk Tenable Cloud Security Skill). This is product and industry news with no incident, vulnerability or active threat described. It is relevant to defenders assessing supply chain risk when deploying agentic AI in SOC workflows.
Technical Details
- Stage 1: Automated screening with the skills-inspection engine in Tenable One AI Exposure. It parses agent instructions, authorized tools and data access, and flags prompt injection, jailbreaks, hidden instructions, hardcoded secrets, PII exposure and sensitive data access.
- Stage 2: Frontier assessment with OpenAI GPT Cyber models via Daybreak. Daybreak Blue supports source code review and dual-use components. Daybreak Red is reserved for higher-risk submissions that need exploit validation and adversarial testing.
- Stage 3: Tenable researchers validate findings and run the component in a clean environment to verify runtime behavior. They produce an auditable record covering provenance, threat model, source review and runtime verification.
- The review covers 15 issue classes across the model, application and infrastructure layers. Examples are prompt injection, excessive agency, memory poisoning, secrets handling, SSRF, path traversal, command execution and supply chain risks such as typosquatting and unpinned packages.
- Tenable reports that SOC-Hunter, a Claude Code skill, cut hunt times by 75% (from 4-6 hours to 45-90 minutes).
Remediation
- Prefer vetted listings, and review the public GitHub source of any community agent, skill or MCP server before production use.
- Assess the permissions, credentials and tool access of AI agents already running in your environment, for example with the skills inspection in Tenable One AI Exposure.