open-source · since 2026
harness-eval
Red Hat Community AI Tools
An Apache-2.0 linter and security auditor for AI coding-agent harnesses. It auto-discovers instructions, skills, hooks, subagents and MCP configurations, builds a cross-component graph, and checks structural, quality, cost and security risks with deterministic rules.
AI agentsdeveloper toolssecurityMCPevaluation
desk notes
Verified 2026-09-20 in the canonical repository, README, LICENSE and GitHub release API. Supports Claude Code, Cursor, Windsurf, Cline, Copilot, Gemini CLI, OpenCode and Codex CLI setups; its current README documents 108 deterministic rules and cross-tool analysis. Available as a CLI, GitHub Action, Tekton Task, plugin/commands and pre-commit hook. Latest verified release: v7.16.0, published 2026-09-16.