A personal, curated library of skills for AI agents.
This is not a bulk collection. Every skill here is one I personally use, have read in full, and have vetted before publishing. Quality and honesty over quantity.
Skills follow the Agent Skills open format and are compatible with any agent that supports the standard — Claude Code, Cursor, GitHub Copilot, Codex, Gemini CLI, and others.
npx skills@latest add jp-cruz/agent-skills/<skill-name>Or install globally to share across all your projects:
npx skills@latest add jp-cruz/agent-skills/<skill-name> --global| Skill | Description | Version | Install |
|---|---|---|---|
| row-bot-docker-setup | Production-ready Docker Compose setup for Row-Bot (formerly Thoth) with lessons learned from bare-metal migrations. Cross-platform support (macOS, Windows, Linux), smart defaults, one-command setup. Includes migration guide for existing Row-Bot users, automated environment detection, multi-provider LLM integration, security hardening, persistent volumes, and comprehensive troubleshooting. Works in any terminal. | v0.5.0 | npx skills@latest add jp-cruz/agent-skills/row-bot-docker-setup |
| legionforge-claude-obsidian | Complete session context initialization system for Claude Code and Desktop. Three-layer context loading (automatic, on-demand via MCP, selective by domain). Provider-independent, crash-safe, cross-platform. Includes setup wizard, verification protocol, and recovery tools. | v0.1.0-alpha | npx skills@latest add jp-cruz/agent-skills/legionforge-claude-obsidian |
| fact-checker | Systematic fact verification with temporal reasoning and search result validity checks. Flags claims that fall outside model knowledge cutoff with recommended follow-up paths. Forked from Shubhamsaboo/awesome-llm-apps. | v1.1.0 | npx skills@latest add jp-cruz/agent-skills/fact-checker |
Skills are executable instructions injected into agent context. A poorly written or malicious skill can cause unintended agent behavior — including prompt injection, tool misuse, or data leakage.
Every skill in this repo has been:
- Read in full — no blind forks or bulk imports
- Reviewed for prompt injection patterns — instructions that could hijack agent behavior
- Checked for scope creep — skills should do one thing and state it clearly
- Attribution verified — forked skills credit the original author and link to the source
If you spot an issue with any skill, please open an issue. I'd rather know.
Skills are provided as-is under MIT license. Review before installing. I am not responsible for how your agent interprets or applies these instructions in your specific environment.
Have a skill you think belongs here? Open an issue and make the case. I'm not accepting PRs, but genuine suggestions are welcome. I'll evaluate, audit, and credit you if I add it. Thank you! - Jp Cruz
MIT — see individual skill directories for attribution on forked work.