Meta Joins Coalition for Secure AI as Premier Sponsor to Advance Industry Security Standards
February 3, 2026
Cisco Donates Project CodeGuard to Coalition for Secure AI
February 9, 2026
By Omar Santos, CoSAI Project Governing Board Co-Chair
In October 2025, Cisco open-sourced Project CodeGuard, a new framework that embeds secure-by-default practices directly into AI coding agent workflows. I am honored and excited to announce that Cisco has donated Project CodeGuard to the Coalition for Secure AI (CoSAI). This contribution marks a significant milestone in our collective mission to advance AI security through open collaboration.
What is Project CodeGuard?
Project CodeGuard is a model-agnostic security framework designed to prevent vulnerabilities before they’re written. Rather than relying on post-generation scanning, it provides AI coding agents with comprehensive security skills and rules that guide them to generate secure code from the first keystroke. However, this framework is also very powerful for code review and remediation.
AI coding agents have transformed how software is built, but speed without security creates risk. Project CodeGuard provides comprehensive security agent skills and rules that guide AI assistants to generate secure code. Project CodeGuard can be used with all leading AI coding agents, including Claude Code, Codex, Antigravity, Cursor, Windsurf, GitHub Copilot, and others.
Why This Donation Matters
Cisco originally developed Project CodeGuard to address internal security challenges with AI-assisted development. Recognizing that the entire industry faces these same challenges, Cisco open-sourced the project in October 2025. Now, by donating Project CodeGuard to CoSAI, Cisco is taking the next step: ensuring the framework evolves through community governance.
This donation reflects a core principle that CoSAI and its members share: AI security is a collective challenge that demands collective solutions. No single organization can address the breadth of security concerns that we face nowadays. By bringing Project CodeGuard under CoSAI stewardship, we can:
- Accelerate innovation through contributions from diverse security experts across industries
- Ensure neutrality as the framework evolves to support all AI coding platforms
- Build trust through transparent, community-driven governance
- Scale adoption by leveraging CoSAI’s network of leading technology organizations and security experts
What are AI Coding Agent Skills and Rules?
Modern AI coding agents can be customized through rules and extended through agent skills. These are two complementary mechanisms that Project CodeGuard leverages to embed security throughout the development workflow.
Rules are plain-text instruction files (typically markdown) that guide AI agents on how to generate, review, and modify code. They function as persistent prompts that shape agent behavior. Rules can be scoped globally, per-project, or activated conditionally based on file patterns. Project CodeGuard provides automated translators that convert a single rule definition into different platform’s native format.
Agent Skills are an emerging open standard for extending AI agent capabilities with reusable knowledge components. Skills enable dynamic context discovery, allowing agents to pull relevant expertise on demand rather than loading everything upfront. This makes them ideal for specialized domains like security, where agents can invoke security-focused skills when working with authentication code, cryptographic operations, or input validation.
Project CodeGuard distributes security guidance through both mechanisms: rules that are always active (preventing hardcoded credentials, enforcing secure cryptographic algorithms) and skills that agents can invoke when deeper security expertise is needed (certificate validation, threat modeling for specific vulnerability classes). This is a powerful combination that allows Project CodeGuard to be used with all leading AI coding agents, including Claude Code, Codex, Antigravity, Cursor, Windsurf, GitHub Copilot, and others.
Get Involved
Project CodeGuard and CoSAI thrive on community collaboration. Visit the Project CodeGuard GitHub repository to get started. Join us to make secure AI coding the default, not the exception. We look forward to your contributions!
