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 collection provides a wide array of structured skill definitions designed to refine how AI agents handle specific professional tasks. By defining precise roles and operational boundaries, it helps builders move from general-purpose prompting to specialized agentic behaviors.
The repository supports a diverse range of activities, from conducting automated accessibility audits and setting up A/B tests to designing complex memory systems and orchestrating agent workflows.
WHO IT'S FOR
AI agent / automation builders
designing complex agentic workflows and memory systems
Frontend / accessibility engineers
performing automated a11y audits on web interfaces
Digital marketing specialists
optimizing ad creatives and setting up AB tests
Software quality assurance engineers
implementing adversarial reviews and agent decision tracking
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.
Installs: The complete repository skill library, installed to all agents supported by Agent Skills CLI. · Claude Code, Cursor, VS Code/Copilot, Goose, Amp, Codex, Letta, OpenCode
Before you start
An environment with npx available.
Run this terminal command to install the complete skill library to every supported agent detected by Agent Skills CLI.
Installs: The complete repository skill library mirrored into .gemini/skills/. · Gemini CLI
Before you start
Git.
A POSIX-compatible shell capable of running the supplied setup script.
Platform: The exact commands shown are suitable for POSIX-style environments. The repository separately warns that Windows clones require Developer Mode, symlink support, and PYTHONUTF8=1.
Run from: Start in the parent directory where the claude-skills repository should be created; run the setup script from the cloned claude-skills repository root.
A POSIX-compatible shell capable of running the supplied setup script.
Platform: The exact commands shown are suitable for POSIX-style environments. The repository separately warns that Windows clones require Developer Mode, symlink support, and PYTHONUTF8=1.
Run from: Start in the parent directory where the claude-skills repository should be created; run the setup script from the cloned claude-skills repository root.
Installs: Only the repository's engineering-team directory, copied into the user's OpenClaw skills directory. · OpenClaw
Before you start
Git.
An OpenClaw installation using ~/.openclaw/skills/.
Platform: The exact commands shown use a POSIX-style home-directory path. The repository separately warns that Windows clones require Developer Mode, symlink support, and PYTHONUTF8=1.
Run from: Run both terminal commands from the parent directory where the claude-skills repository should be created.
Stress-test the first unresolved decision in a plan
Uses grill-me
Replace <path-to-plan> with the path to the plan or design to review, or provide the plan directly.
A codebase or workspace to inspect if the plan concerns an existing project.
Using the grill-me skill, stress-test the plan at <path-to-plan>. Explore the codebase first wherever it can answer a question. Then ask me one question at a time, include your recommended answer and a one-sentence rationale, and begin with the first unresolved dependency. Stop after the first question so I can respond.
Replace <team-name> with the responsible team or owner.
Replace <environment> with the target environment, such as staging.
Using the runbook-generator skill, create a draft baseline runbook for <service-name>, owned by <team-name>, in <environment>. Include start, stop, health-check, rollback, escalation, and incident-handling placeholders. Identify the service-specific commands and URLs I still need to supply; do not deploy or run production actions.
Replace <level> with the role level, such as mid or senior.
Optional: provide existing competencies, time constraints, or the current interview process.
Using the interview-system-designer skill, draft a baseline interview loop for a <role> at <level>. For each round, give its focus, suggested timing, a small sample question set, and the scoring or calibration guidance. Include bias-control checks and keep the result as a draft for panel-lead review.
Claude
Codex
Cursor
Gemini CLI
OpenClaw
Copilot
Windsurf
Amp
OpenCode
Hermes
Antigravity
Goose
Kilo Code
Mistral Vibe
Cline