Boot.gs Skills

Agent Skills for boot.gs – a NestJS/Spring Boot-style framework for Google Apps Script (decorators, dependency injection, routing, validation) — plus a set of skills for plain Google Apps Script development that don't require boot.gs at all.

Each skill is a self-contained directory under skills/ with a SKILL.md file and, where useful, scripts/, references/, or assets/. The format follows the open Agent Skills specification, so every skill here works with any agent that implements it – not just Claude. An llms.txt at the repo root gives agents and crawlers a short index of what's here.

Skills

Boot.gs

apps-script (framework-agnostic)

Installation

Skills run with the full permissions of whatever agent loads them. Read a skill's SKILL.md and any bundled scripts/ before installing it, the same way you would before running someone else's code.

Claude Code

Register this repository as a plugin marketplace:

/plugin marketplace add bootgs/skills

Install just the boot.gs skills, just the framework-agnostic ones, or both:

/plugin install bootgs@bootgs-skills
/plugin install apps-script@bootgs-skills

Gemini CLI

Install every skill in one extension, discovered automatically from the root gemini-extension.json and skills/:

gemini extensions install https://github.com/bootgs/skills

To install just the bootgs skills or just the framework-agnostic ones, this repository also ships a Gemini CLI marketplace manifest (.agents/plugins/marketplace.json, following the pattern used by google/skills), with the same two groups packaged as standalone extensions under plugins/. Installing a single group from a URL requires cloning first, since gemini extensions install only reads a gemini-extension.json from the exact path given:

git clone https://github.com/bootgs/skills
gemini extensions install ./skills/plugins/bootgs
gemini extensions install ./skills/plugins/apps-script

plugins/bootgs/skills and plugins/apps-script/skills are symlinks into skills/ — content isn't duplicated.

Any agent, via npx skills

Skills also install with the skills CLI, which supports Claude Code, Cursor, Codex, Gemini CLI, and dozens of other agents:

npx skills add bootgs/skills --skill bootgs-quickstart

Try it

Once a skill is installed, prompt the agent in plain language — it decides on its own when a skill applies:

• Set up a new boot.gs project for a Sheets add-on.
• Add a page/limit query parameter to this boot.gs controller, validated and bounded.
• Build a client for my boot.gs backend that works inside a Sheets sidebar.
• Regenerate the OpenAPI spec for this boot.gs project and check it's not stale.
• Why does this onEdit trigger work for me but not for other editors?
• This add-on keeps failing Marketplace review — check it against the current checklist.

Development

Managing skills

package.json wraps the skills CLI so the flags don't need to be memorized:

npm run skills:init             # scaffold a new SKILL.md
npm run skills:add -- <source>  # install a skill from a source
npm run skills:list             # list installed skills
npm run skills:find             # search for skills
npm run skills:use              # preview a skill without installing it
npm run skills:update           # update installed skills
npm run skills:remove           # remove an installed skill

skills:add always targets --agent '*': everything installs into .agents/skills/ as the canonical copy, and the CLI symlinks it into whichever other agent directories already exist in the project — there's no hardcoded agent list to keep in sync, since a clone of this repo may be used with agents we don't know about in advance.

This repository's own tooling — the skill-creator skill, used to draft and iterate on the skills above — lives under .agents/skills/, installed the same way:

npm run skills:add -- anthropics/skills --skill skill-creator

Reviewing skills

.claude/agents/skill-reviewer.md is a fresh-context reviewer for a SKILL.md and its bundled scripts//references//assets/. It doesn't author or edit — it checks description-trigger accuracy against sibling skills, structural conventions against template/SKILL.md.template, progressive disclosure, freshness handling (does a skill stating versioned third-party facts ship a check script or tell the reader to verify against the source?), and whether the ## Verification checklist is actually checkable. Run it in Claude Code after drafting or editing a skill, before publishing a new one:

Use the skill-reviewer agent on skills/<name>/SKILL.md

Support

Found a skill that's wrong, out of date, or misfiring? Open an issue in the issue tracker.

Contributing

Issues and pull requests are welcome — a corrected gotcha, a new skill, or a report that a live-fetched script broke because Google or the bootgs API changed shape are all useful.

License

Apache License 2.0