fencing token
Subscribe

Issues · · DESIGN-REVIEW REFERENCES

Six free agent-security references for your next design review

New specs and references from OWASP, CoSAI, MITRE ATLAS, and the MCP project, each mapped to a question you can ask in a review.

Six free references for your next agent review.

Sources

  1. StandardOWASP GenAI Red Teaming Manual v1.0 (Oct 20)

    An eight-phase method, from planning to remediation, that covers agents. genai.owasp.org

  2. StandardCoSAI: Zero Trust for AI Systems (Sep 16)

    Authorization can't be delegated to the model; bind sessions to the user and issue short-lived, scoped tokens. coalitionforsecureai.org

  3. CodeMITRE ATLAS v2026.09 (Sep 15)

    New agent techniques, such as probing trigger channels and discovering runtime capabilities. github.com

  4. StandardOWASP Agent Control Standard (Sep 1)

    Runtime hooks and portable policy across agent frameworks. Early stage (v0.1). genai.owasp.org

  5. StandardCoSAI: MCP Security v2.0 (Aug 12)

    Four assurance levels across eight security dimensions. coalitionforsecureai.org

  6. StandardMCP spec, 2026-07-28 revision

    Removes protocol-level sessions; clients must validate the issuer in authorization responses when present. modelcontextprotocol.io

Part of the Fencing Token roadmap. Found an error? Open an issue.