Skills Manager

Catalog of command snippets and prompt templates for RobOS apps, with custom entries and installable packs from GitHub.

packages/skills-manager

In this app a “skill” is an entry with a name, category, tags and either a shell command (for example node packages/robos-graph/bin/kgraph-cli.js search "$QUERY") or a system prompt. The catalog feeds RobOS apps such as AI Prompt. It is separate from the SKILL.md agent skills that Claude Code and similar tools load; for those, see Skills.

Use it

  • My Skills lists the built-in skills and your own, with a search box (name, description or tag) and category filter. The built-ins are command snippets in categories such as Knowledge Graph, File Operations, Process Management, Git, Network, Docker, System, Package Management, Text Processing, Security and Development, defined in packages/skills-manager/skills-data.js.
  • + New Skill opens a form: Name, Category, Description, Command / Script and Tags (comma-separated). Custom skills are saved in ~/.config/robos/skills.json; built-ins are not stored there.
  • Open AI Prompt launches the AI Prompt app from /usr/local/share/robos/ai-prompt.
  • Skill Packs lists three featured packs. Select one to browse its patterns, filter them, tick the ones you want, and use Install Selected or Install All. Installed patterns become custom skills with the pack’s system prompt as their content.
Pack Source Needs
Fabric Patterns github.com/danielmiessler/fabric (data/patterns) Network access, and git if you click Clone Repo (a shallow clone to ~/.config/robos/skill-packs/fabric)
Awesome Cursor Rules github.com/PatrickJS/awesome-cursorrules (rules) Network access, and git to clone
SDLC Essentials Built into the app (code review, PR writing, testing, documentation and similar prompts) Nothing

Browsing the two GitHub packs uses the GitHub contents API unless the repository is already cloned. The pack descriptions in the app quote pattern counts and star counts that are written into the source, not fetched, so they may be out of date.

Imported entries are named with a category prefix (Fabric:, CursorRules: or SDLC:). An entry that is already installed is skipped.

Related guides

More in AI & Agent Workflows