agent-harness
Install only the AI-coding skills & plugins your tech stack actually needs.
Coding agents keep piling on skills and MCPs you'll never use — it just bloats their context. agent-harness curates by tech stack: of the vetted options, you install only what this project calls for, across Claude Code, Codex, OpenCode & Antigravity. Project scope by default — no global pollution unless you ask.
Why
Pick your stack; install only the vetted skills, plugins, rules & hooks that project needs — not a kitchen sink.
Every catalog asset is ≥1000★ + actively maintained + Docker install-verified (49/58 green), re-checked monthly for star-drift.
Default writes stay inside the project. Existing config is backed up before any change — nothing is silently overwritten.
Vetted ≠ a security audit. The compatibility matrix is generated from real install runs, not a hand-maintained table.
Across four CLIs (honestly labeled)
- Claude Code — first-class: skills, plugins, rules, hooks, 6-gate workflow.
- Codex / Antigravity — native skills + rules layer (
.agents/skills/, AGENTS.md). - OpenCode — rules + command fallbacks (no native skill concept upstream).
What "vetted" means: ≥1000 GitHub stars, active maintenance, and a real install in an isolated Docker container — not a line-by-line security audit or a prompt-injection scan of asset contents. npm/npx assets are version-pinned; plugin/skill assets resolve to upstream HEAD. Treat installed assets like any third-party dependency.