Tags: docker/cagent-action
Tags
Initial release of cagent-action A GitHub Action for running CAgent AI agents in CI/CD workflows with built-in security hardening and comprehensive prompt injection protection. Key Features: - Download and execute CAgent agents from Docker Hub or local paths - Optional MCP Gateway integration for containerized MCP servers - Multi-provider AI model support (Anthropic, OpenAI, Google) - Interactive TUI mode and YOLO mode for auto-approval - Binary caching for faster workflow execution Security Features: - Input sanitization with 45 HIGH-RISK and 4 MEDIUM-RISK pattern detection - Comment removal from diffs to prevent hidden prompt injection - Output scanning for leaked secrets (API keys, tokens) - Automatic incident response with security issue creation - Authorization checks for PR author permissions - Comprehensive test suite (13 security + 6 exploit tests) Co-authored-by: Derek Misler <[email protected]> Co-authored-by: Jean-Laurent de Morlhon <[email protected]>