feiskyer/claude-code-settings
Curated skills, sub-agents, and config templates that supercharge Claude Code — research, image gen, GitHub automation & more.
Claude Code Skills & Agents – What It Is
feiskyer/claude-code-settings is a collection of ready‑to‑install skills (modular capability plug‑ins) and sub‑agents for Claude Code, Anthropic’s AI‑assisted coding environment. The repo bundles configuration templates, installation scripts, and a catalog of 15+ skills that let Claude perform tasks such as brainstorming, deep research, GitHub issue fixing, PR review, image generation (via Gemini or OpenAI), You‑Tube transcription, translation, and more.
Quick start (official way)
- Add the plugin from Claude’s marketplace
/plugin marketplace add feiskyer/claude-code-settings /plugin install claude-code-settings # installs all skills - Or use the CLI version:
claude plugin marketplace add feiskyer/claude-code-settings claude plugin install claude-code-settings@claude-code-settings - The plugin installs the full skill set; individual skills can be added later with
npx skills.
Main capabilities (skill list)
| Skill | Core function |
|---|---|
| brainstorming | Interactive design‑idea generation and spec drafting |
| codex‑skill | Non‑interactive automation that hands tasks to OpenAI Codex (code generation, review, planning) |
| deep‑research | Multi‑agent parallel research workflow that splits a query, runs sub‑processes, and assembles a polished report |
| github‑fix‑issue | End‑to‑end GitHub Issue remediation (branch, code, test, PR) |
| github‑review‑pr | Multi‑angle PR review with security checks, confidence scoring, and auto‑LGTM |
| nanobanana‑skill | Image generation/editing via Google Gemini |
| gpt‑image‑skill | Image generation/editing via OpenAI GPT‑Image models |
| youtube‑transcribe‑skill | Pull subtitles/transcripts from YouTube videos |
| translate | High‑quality English/Japanese → Chinese technical translation |
| grill‑me | Intensive design‑review interview style questioning |
| handoff | Produce a concise hand‑off document for the next AI agent |
| skill‑creator | Scaffold, iterate, and benchmark new Claude skills |
| others | nanobanana‑skill, gpt‑image‑skill, translate, skill‑creator, etc. |
Configuration templates
The repo ships 10 pre‑filled settings.json files for different model providers (GitHub Copilot, LiteLLM, DeepSeek, Qwen, SiliconFlow, Vertex AI, Azure, MiniMax, OpenRouter, etc.). Swap the file that matches your backend and adjust ANTHROPIC_BASE_URL, ANTHROPIC_API_KEY, and model names.
Installation alternatives
npx skills– install a single skill or the whole set (may lag behind the latest version). Example:npx -y skills add --all feiskyer/claude-code-settings- Manual clone – replace
~/.claudewith the repo, then start a Copilot‑gateway proxy (http://localhost:4141) or any LiteLLM‑compatible proxy.
Typical workflow examples
- Research a new technology – invoke
/deep-research <topic>; Claude spawns parallel agents, gathers web data, and returns a markdown report in.research/<topic>/final_report.md. - Fix a GitHub issue – run
/github-fix-issue <issue‑url>; Claude creates a branch, writes code, runs tests, and opens a PR. - Generate UI mock‑ups – use
/nanobanana-skillor/gpt-image-skillwith a prompt; the skill calls Gemini or OpenAI image models and returns image files. - Hand off a session – after a long coding session,
/handoff <next‑goal>produces a sanitized hand‑off markdown that can be fed to another agent.
Who might use this?
- Developers who already work with Claude Code and want richer, plug‑and‑play capabilities without writing their own skill code.
- Teams looking to automate repetitive GitHub workflows (issue fixing, PR review) via an LLM.
- Researchers needing structured, multi‑agent web research directly inside Claude.
- Content creators who want on‑the‑fly translation or image generation integrated with their coding workflow.
Getting help / contributing
- FAQ section in the README covers skill loading, model switching, permission modes, and proxy authentication.
- Contributions are welcomed via issues and pull requests; see
CONTRIBUTING.mdfor dev/testing steps. - The repository also provides helper scripts, e.g.,
update-cc-plugins.shto refresh all installed plugins.
License
MIT – free to use, modify, and distribute.
Related
- Project
- Project
- Project
- Project