emt-brain 1.3.1

EMT Brain — sync skills, workflows, agents into Claude Code

emt-brain

EMT Brain — populate skills, workflows, and agents from the EMT repository into Claude Code (or GitHub Copilot).

Install

# From wheel (recommended)
pip install dist/emt_brain-0.6.0-py3-none-any.whl

# From source
pip install -e tools/emt-brain

# With uv
uv pip install dist/emt_brain-0.6.0-py3-none-any.whl

Quick start

# One-shot full setup (MCP servers + 3rd-party plugins + EMT skills/agents)
emt-brain init

# Or step by step:
# 1. Install MCP servers (context-mode)
emt-brain 3rd-mcp-setup

# 2. Install 3rd-party plugins (e.g. graphify)
emt-brain 3rd-plugins-setup

# 3. Sync default bundle (master + utilities + anthropics + context-engineering)
#    + all workflows + all agents
emt-brain framework-sync

# 4. See available workflows
emt-brain workflow-list

# 5. Scope to a specific workflow and its dep closure
emt-brain framework-sync --workflow emt-wf-ip-planning

# Preview without writing
emt-brain framework-sync --dry-run

# Show onboarding guide
emt-brain help

Commands

emt-brain init

Full one-shot setup that runs three steps in order:

1. 3rd-mcp-setup — install Context Mode MCP server + Memory (npx, optional)
2. 3rd-plugins-setup — install 3rd-party skill/agent plugins (e.g. graphify)
3. framework-sync — sync EMT default bundle + all workflows + agents

emt-brain init                        # interactive full setup
emt-brain init --yes                  # non-interactive (CI)
emt-brain init --dry-run              # preview without writing
emt-brain init --scope project        # install into project .claude/
emt-brain init --skip-mcp            # skip MCP servers step
emt-brain init --skip-plugins        # skip 3rd-party plugins step
emt-brain init --skip-sync           # skip framework-sync step

emt-brain init-cowork

Install document & productivity skills to local Claude Code. A focused variant of init that syncs only the 10 co-work skills and installs their runtime dependencies (npm/pip). Useful for users who want document generation tools without the full EMT framework.

Skills installed (10):

Source	Skills
FPT utilities	emt-sk-brand-guide, emt-sk-docx, emt-sk-jp-trans, emt-sk-pdf, emt-sk-pptx
Anthropic community	docx, pdf, pptx, xlsx, doc-coauthoring

Runtime deps auto-installed:

• emt-sk-docx — Node.js deps (npm install) if node_modules/ is missing
• emt-sk-pdf — Python deps (pip install -r requirements.txt): reportlab, PyYAML, pypdf, Pillow

emt-brain init-cowork                  # install co-work skills + deps to ~/.claude/skills/
emt-brain init-cowork --dry-run        # preview without writing
emt-brain init-cowork --scope project  # install into project .claude/
emt-brain init-cowork --skip-deps      # skip npm/pip dependency installation
emt-brain init-cowork --force          # overwrite locally modified files

emt-brain 3rd-mcp-setup

Install and activate MCP servers into Claude Code's settings.json. Checks if each server is already configured — skips if present, adds if missing. Restart Claude Code after running to activate.

emt-brain 3rd-mcp-setup                          # write to ~/.claude/settings.json (global)
emt-brain 3rd-mcp-setup --scope project          # write to .claude/settings.json (project)
emt-brain 3rd-mcp-setup --force                  # reinstall local servers (re-clone + npm install)
emt-brain 3rd-mcp-setup --dry-run                # preview without writing

Servers installed:

Server	Mode	Package / Source
memory	npx	@modelcontextprotocol/server-memory
context-mode	local	github.com/mksglu/context-mode

npx servers require Node.js. Registered as npx -y <package> — no local storage.

local servers (e.g. context-mode) are cloned and installed at runtime:

1. git clone --depth=1 <source> → ~/Library/Application Support/emt-brain/mcp-servers/<name>/
2. npm install --prefer-offline
3. npm run build if package.json has a build script
4. Entrypoint auto-detected: bin → main → index.js from package.json
5. Registered in settings.json as { "command": "node", "args": ["/path/to/entry.js"] }

local mode requires git + npm on PATH in addition to Node.js.

emt-brain 3rd-plugins-setup

Install 3rd-party skill/agent plugins from public GitHub repos. Plugins are cloned with git clone --depth=1 and copied into .claude/skills/ or .claude/agents/. No authentication required.

MCP servers (like context-mode) are managed by 3rd-mcp-setup, not here.

emt-brain 3rd-plugins-setup                     # install all missing plugins
emt-brain 3rd-plugins-setup --list              # show status, do not install
emt-brain 3rd-plugins-setup --plugin graphify   # install one plugin
emt-brain 3rd-plugins-setup --force             # reinstall even if present
emt-brain 3rd-plugins-setup --dry-run           # preview without writing
emt-brain 3rd-plugins-setup --scope project     # install into .claude/ (project)

Available plugins:

Plugin	Type	Source	Install path
graphify	skill	github.com/safishamsi/graphify	~/.claude/skills/graphify/

Requires git on PATH.

emt-brain framework-sync

Populate skills, workflows, and agents in one shot.

emt-brain framework-sync                                 # default bundle + all
emt-brain framework-sync --workflow emt-wf-ip-planning  # scope to workflow deps
emt-brain framework-sync --dry-run                      # preview only
emt-brain framework-sync --force                        # overwrite local modifications
emt-brain framework-sync --scope project                # project-local (.claude/skills/)
emt-brain framework-sync --scope global                 # global (~/.claude/skills/)

Default bundle (flezi-emt/bundles/default.bundle.yaml) includes:

• Master skills (10) — EMT meta-tooling
• Utility skills (5) — docs, translations, branding
• Community: Anthropics (7 curated)
• Context-engineering skills (7 — context engineering foundation per IDEA-001)

Workflow sync (--workflow <id>) resolves the dep closure:

• Target workflow + transitively consumed workflows
• All agents referenced by those workflows
• All skills used by those agents
• Plus the default bundle (always included)

emt-brain workflow-list

List all available EMT workflows.

emt-brain workflow-list           # summary table
emt-brain workflow-list --full    # include per-workflow agent list
emt-brain workflow-list --json    # raw JSON

emt-brain help

Show categorised command overview and getting-started guide.

emt-brain status

Show current configuration (repo, branch, provider, scope) and drift count.

emt-brain doctor

Run environment diagnostics: Python version, git, Azure auth, repo reachability, destination writable, state file.

emt-brain reset

Delete CLI config and state. Already-synced files are not removed.

Authentication

emt-brain uses azure-identity. On first run it prompts for device-code login if the Azure CLI credential is not cached. Subsequent runs reuse the cached credential.

Project config

Place a .flezi-emt/config.yaml at your project root to configure emt-brain per project:

profile: default            # bundle to sync
provider: claude            # claude | copilot
scope: project              # global | project

CLI flags override project config; project config overrides global config.

Location	Purpose
Global config	~/.config/emt-brain/config.json — repo URL, branch, provider, scope
Project config	.flezi-emt/config.yaml — profile, filter_tags, provider, scope
State	~/.local/share/emt-brain/state.json — managed files, last sync
Cache	~/.cache/emt-brain/repo-cache/ — cloned remote repo

Run emt-brain doctor to see active paths.

Post-sync: Claude Code setup

After syncing, in your project run /emt-brain-help setup-project in Claude Code to:

1. Initialize CLAUDE.md (if absent)
2. Add horizontal-protocol guidelines (IDEA-001 four-pillar session protocol)
3. Add workflow-usage guidelines for each installed workflow