Stats
Actions
Available In
Tags
Plugin
scoping
Turn raw project inputs (briefs, transcripts, screenshots, voice notes, loose prompts) into a complete, staged scoping package — scope.md, screens.md, phased task cards, and API docs.
README
claude-skills
A small collection of custom Claude Code skills, packaged as a plugin marketplace so you can install them with a couple of slash commands.
| Skill | What it does |
|---|---|
| scoping | Turns raw project inputs (briefs, meeting transcripts, screenshots, voice notes, loose prompts) into a complete, staged scoping package — scope.md, screens.md, phased task cards, and API docs. |
| prompt-engineer | Builds production-grade prompts from a short brief: a quick discovery interview, then a fully-scaffolded prompt (role, context, task, inputs, instructions, output spec, examples, optional self-verification). |
| autonomous-work | Runs a full spec → plan → execute pipeline with minimal check-ins. Requires the superpowers plugin — see Requirements below. |
Install
In Claude Code, add this marketplace once:
/plugin marketplace add Mike-Ab/claude-skills
Then install whichever skills you want (à la carte):
/plugin install scoping@claude-skills
/plugin install prompt-engineer@claude-skills
/plugin install autonomous-work@claude-skills
That's it — the skills are now available. They activate automatically when your request matches the skill, or you can describe the task directly (e.g. "scope this project", "write me a prompt for…", "go autonomous").
To update later:
/plugin marketplace update claude-skills
Requirements for autonomous-work
autonomous-work is the most opinionated skill and has two dependencies:
-
The superpowers plugin. It orchestrates superpowers skills (brainstorming, writing-plans, subagent-driven-development, requesting-code-review, etc.). Without superpowers installed, the pipeline's review and planning phases won't have skills to call. Install superpowers first.
-
Session-continuity conventions. The skill writes and reads two files to make long, multi-session work resumable:
progress.md— an append-only log; one## Session N — YYYY-MM-DDentry per session.handover.md— a snapshot, fully rewritten at session end, that lets a fresh agent resume cold.
It also expects an optional
docs/<feature>.mdconvention for feature documentation. These conventions live in the author's globalCLAUDE.md. If you want the same behaviour, add equivalent rules to your ownCLAUDE.md, or just let the skill create the files as it goes — it will.
scoping and prompt-engineer have no dependencies and work standalone.
Manual install (no marketplace)
If you'd rather not use the plugin system, copy any skill folder straight into your skills directory:
git clone https://github.com/Mike-Ab/claude-skills.git
cp -R claude-skills/plugins/scoping/skills/scoping ~/.claude/skills/
Restart Claude Code and the skill will be picked up.
License
MIT © Mike Ab
Stats
Version1.0.0
Stars0
MaintenanceGood
LicenseMIT
Last Commit
Added
Available In
Tags
Categories
