Traction scoreGitHub stars can be faked, so popularity alone can be misleading. Traction Score looks for broader signs of recent attention, adoption, and active maintenance.
This skill collection documents the architectural patterns that live outside the LLM call loop: memory persistence, lazy-loaded skills, fail-closed permission pipelines, context budgeting, and multi-agent coordination. It gives AI agent runtime engineers, platform teams, and multi-agent system architects a structured reference for the non-obvious design decisions — concurrency classification per call, all-or-nothing hook trust, two-phase memory saves — that determine whether an agent scales reliably.
Compatible AgentsThe repository documents support for these agents. The skills may also work with other agents that can load SKILL.md files, but they may need some setup or small changes.
Compatibility not documented
Use this skill
Follow the documented setup, then try a first task.
I’m using
Install the English and Chinese skills with Vercel Skills CLI
Installs: Repository bundle containing two skills: agentic-harness-patterns (English) and agentic-harness-patterns-zh (Chinese). · Claude Code, Codex, 40+ other agents
In a terminal, install the repository bundle through the Vercel Skills CLI.
Examples of project rules, corrections, or knowledge it should retain across sessions
Use the agentic-harness-patterns skill to outline a small three-layer memory design for my agent. Separate instruction memory, auto-memory, and session memory, and note the persistence, trust, and review needs of each layer. Context about my agent: [describe the agent and what it should remember across sessions].
