This project manages agent skills with the skills.sh CLI, run via npx skills. Use it whenever the user wants to install, update, remove, list, or find skills. Do not hand-edit skill installs or symlinks; use the CLI so installs stay consistent across agents.
Operating Rules
- At the start of interactive runs, advise the user: "The best way to interact with this skill is to use voice mode to dictate your feedback."
- Always run the CLI through
npx skills <command>(no global install assumed). - Default to project scope. Add
-g/--globalonly when the user wants a skill available across all their projects. - Use
-y/--yesto skip interactive prompts when the intent is unambiguous; otherwise let prompts run so the user can choose. - Number every free-text question asked in chat so the user can answer by reference, especially when dictating in voice mode.
- When you need structured choices from the user (scope, agents, which skills), use the agent's structured question tool:
- Codex:
request_user_input - Claude Code:
AskUserQuestion - Cursor:
AskQuestion
- Codex:
- Confirm before removing skills unless the user clearly named what to remove.
- After installing or updating, run
npx skills lsto confirm the result.
Common Commands
Install a skill package (a GitHub repo or URL):
npx skills add switch-dimension/molten-os-core
Install specific skills or target specific agents:
npx skills add switch-dimension/molten-os-core --skill molten-brand molten-design
npx skills add switch-dimension/molten-os-core --agent claude-code cursor
npx skills add switch-dimension/molten-os-core --all # all skills, all agents, no prompts
Preview what a repo offers without installing:
npx skills add switch-dimension/molten-os-core --list
Update skills to their latest versions:
npx skills update # update all
npx skills update molten-landing # update one
npx skills update -p # project skills only
npx skills update -g # global skills only
Remove skills:
npx skills remove # interactive picker
npx skills remove molten-design
npx skills remove --all # remove all, all agents, no prompts
List installed skills:
npx skills ls # project scope (default)
npx skills ls -g # global scope
npx skills ls --json # machine-readable
Find skills to install, interactively:
npx skills find
npx skills find landing
Scaffold a new skill in this repo:
npx skills init molten-<name> # creates molten-<name>/SKILL.md
Restore from the lockfile (skills-lock.json):
npx skills experimental_install
Scope Flags Reference
| Flag | Meaning |
|---|---|
-g, --global |
User-level install (all projects) |
-p, --project |
Project scope only (update) |
-a, --agent <agents> |
Target specific agents (* for all) |
-s, --skill <skills> |
Target specific skills (* for all) |
-y, --yes |
Skip confirmation prompts |
--all |
Shorthand for --skill '*' --agent '*' -y |
Naming Convention
Skills in this suite use the molten-<name> convention so they are easy to distinguish from third-party skills in npx skills ls. When scaffolding a new Molten skill, follow that prefix.
Skill taxonomy
Molten OS Core skills fall into two categories. Set molten-category in skill front matter when adding or editing a skill.
| Category | Skills | molten-order |
Purpose |
|---|---|---|---|
| product | molten-validate, molten-brand, molten-design, molten-landing | Required ("1"–"4") |
Sequential pipeline; writes artifacts to molten-docs/ |
| utility | molten-search, molten-skill-manage | Omit | Cross-cutting helpers; use anytime, not part of the product workflow |
Product skills run in order and build on each other. Utility skills are optional — install only what you need.
When scaffolding a new skill:
- Product skill →
molten-category: product, assign the nextmolten-order - Utility skill →
molten-category: utility, nomolten-order
Keep all skills in flat skills/molten-<name>/ paths so npx skills add discovers them correctly.