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 repository provides a structured approach for AI agent builders to implement and manage custom skills. It enables the creation of specialized behaviors, such as auditing another agent's work or orchestrating complex workflows across codebase-heavy tasks.
By defining discrete skill sets, developers can better control how agents investigate problems, verify changes, and manage token usage during research and coding sessions.
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.
Follow the documented setup, then try a first task.
I’m using
Interactive skills installer
Installs: Selected catalog skills at user or project scope; recommended skills are preselected. The installer offers shared .agents paths and Claude Code's native skills path. · Codex, Pi, Cursor, OpenCode, GitHub Copilot / VS Code, Claude Code
Run this command in a terminal to open the installer.
npx @agent-native/skills@latest add
Choose skills, agent clients, and user-level or project-level scope. Review the offered visual-output settings, managed instruction blocks, and optional integrations for selected skills.
Installs: Selected quick-recap skill at user or project scope, with optional managed instruction guidance. · Codex, Pi, Cursor, OpenCode, GitHub Copilot / VS Code, Claude Code
Run this command in a terminal to select quick-recap without the skill picker, then choose the installation client and scope.
Installs: All skills as a managed, updatable Builder Skills plugin, namespaced under builder-skills. This procedure does not add managed instruction blocks or the PR Visual Recap GitHub Action; Rewind instructions require separate local connection setup before use. · Claude Code
Enter both commands inside a Claude Code agent session.
A work summary stating what is complete and any remaining follow-up or missing user input.
Make sure quick-recap is installed in your agent before trying this task.
Use quick-recap to turn the work summary I provide into one final status line under 100 characters. Choose green for finished work, yellow for a specific non-routine follow-up, or red when blocked on user input, and name any pending item.
An existing .agent-factory/config.yaml authorizing the repositories and sources to inspect.
Read-capable host integrations for the configured pull-request sources.
Make sure factory-human-digest is installed in your agent before trying this task.
Use factory-human-digest for a brief, read-only digest of configured pull requests from the last 7 days that still need human judgment. Report coverage and unavailable or truncated sources, then list the top groups, why each needs a person, and links to the source items. Do not take external actions.
Make sure webmcp is installed in your agent before trying this task.
Use webmcp to open the URL I provide in the host's visible built-in browser. Confirm the page is there and stop without listing tools, inspecting schemas, signing in, or starting app work.
Claude
Codex
Cursor
Copilot
OpenCode