A single skills repository for AI coding agents (Claude Code, Codex CLI, and others), structured Matt-Pocock style: every skill lives at skills/<category>/<name>/SKILL.md, the whole repo is one plugin, and the primary install path is the tool-agnostic npx skills installer.
Skills in this repo install into 70+ agents (Claude Code, Codex, Cursor, Copilot, ...) via the skills CLI:
# Interactive: pick skills + target agents
npx skills@latest add devstefancho/claude-plugins
# List available skills without installing
npx skills add devstefancho/claude-plugins --list
# Install one skill non-interactively
npx skills add devstefancho/claude-plugins --skill writing-specs -a claude-code -y
# Install everything to all detected agents
npx skills add devstefancho/claude-plugins --allPick the skills you want; the installer copies them into your coding agent. Re-run to add or update skills.
Codex CLI: install via the same
npx skills addpath. (If your Codex version doesn't yet support npx-installed skills, clone the repo and point your skills directory atskills/— see Local development.)
| Skill | Description |
|---|---|
| writing-specs | Write and manage spec files with search, conflict detection, and reporting |
| writing-tasks | Decompose specs into persistent task files with a dependency graph and progress |
| writing-flows | Write single-scenario Flow docs with a Mermaid diagram, step branches, and source references |
| implement-with-test | Implement a task with tests; auto-detects the test framework |
| test-commit-push-pr-clean | Branch-safe finish: lint, test, commit, push, open PR, clean worktrees |
| Skill | Description |
|---|---|
| create-team | Create and manage a planner + implementer agent team (create / cleanup / expand) |
| split-work | Split current work into parallel-safe task groups with worktree branches |
| Skill | Description |
|---|---|
| browser-walkthrough | Headed, step-by-step browser walkthrough for iframe/security-heavy sites |
| computer-use-test | Run app test scenarios via Computer Use MCP and report UI/UX feedback |
| ui-prototype-preview | Turn a saved brainstorm idea into a standalone HTML prototype |
| Skill | Description |
|---|---|
| brain-storm | Brainstorm features/improvements from the current codebase (pre-spec ideation) |
| session-resume | Resume a previous Claude Code / Codex session from its JSONL transcript |
| llm-wiki | Maintain an LLM-powered personal wiki from raw sources |
| Skill | Description |
|---|---|
| hermes-runtime | Talk to / control the Hermes companion runtime (chat, run, status, jobs, setup) |
| setup-notification | Install macOS TTS + dialog hooks for Claude Code Stop/Notification events |
Every skill follows the same conventions (inspired by mattpocock/skills):
- Frontmatter
descriptionis two sentences: what it does, thenUse when [explicit triggers]. SKILL.mdstays under ~100 lines — terse, imperative, with phase workflows and checklist gates.- Overflow detail lives in sibling files (
reference.md,templates/,scripts/), linked one level deep.
See the SKILL.md conventions section in AGENTS.md before adding or editing a skill.
.claude-plugin/plugin.json # single plugin manifest — lists every skill
skills/
<category>/
<skill-name>/
SKILL.md # required: name + description frontmatter + instructions
... # optional supporting files: scripts/, commands/, templates/
docs/adr/ # architecture decision records
evals/ # skill eval suites (dev only, not shipped)
A skill is the unit. There are no per-plugin wrappers, no marketplace.json, and no per-tool manifests — multi-agent reach is delegated to the npx skills installer.
git clone https://github.com/devstefancho/claude-plugins.git
cd claude-plugins- Install from your local checkout:
npx skills@latest add ./ - Or load the whole repo as a plugin in Claude Code via
--plugin-dir(seescripts/cldp.sh).
See AGENTS.md for conventions and how to add a skill.
MIT