Agent Team Plugin

devstefancho skills

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.

Install

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 --all

Pick 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 add path. (If your Codex version doesn't yet support npx-installed skills, clone the repo and point your skills directory at skills/ — see Local development.)

Skills

spec-driven

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

agents

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

browser

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

productivity

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

misc

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

Skill style

Every skill follows the same conventions (inspired by mattpocock/skills):

• Frontmatter description is two sentences: what it does, then Use when [explicit triggers].
• SKILL.md stays 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.

Repository structure

.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.

Local development

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 (see scripts/cldp.sh).

See AGENTS.md for conventions and how to add a skill.

License

MIT