tomershahar/agent-harness-kit

agent-harness-kit is a toolkit to make AI agents and new developers productive and consistent in a repository by creating a repeatable "harness"—the files, scripts, and conventions that determine agent reliability. It targets engineering teams and agents/tools (Claude, Copilot, Gemini CLI, etc.) that need durable context, onboarding, and session handoffs. Key features: - Four skills: harness-init (auto-generate a complete harness), harness-audit (score five subsystems: Instructions, Tools, Environment, State, Feedback and produce prioritized fixes), harness-handoff (capture session decisions and blockers), harness-onboard (orient new contributors in minutes). - Generates AGENTS.md, ARCHITECTURE.md, init.sh, feature_list.json, PROGRESS.md and embeds HARNESS-GAP comments where fixes are needed. - Quick install scripts, audit report output, and a 91-test suite for validation. MIT licensed.

このRepoのdescription / GitHub topics / READMEのAI要約に出現したAI関連キーワードをカテゴリ別に表示。各バッジは該当カテゴリの詳細ランキングへリンクします。