Migma

Generate, send, validate, and export AI-powered emails from the terminal; manage contacts, segments, tags, domains, and webhooks with Migma CLI.

Security Scan

The skill is internally consistent with an email CLI: it requires the migma binary and a single MIGMA_API_KEY, and its instructions only use the CLI to generate, validate, export, and send email content (including importing local CSVs) — nothing appears to be requesting unrelated secrets or privileged system access.

Purpose & Capability

Name/description, required binary (migma), required env var (MIGMA_API_KEY), and the declared Node install (@migma/cli) all align with an email/marketing CLI. The commands in SKILL.md (generate, send, validate, export, contacts) match the stated purpose.

Instruction Scope

The instructions are limited to running the migma CLI and operating on files the user explicitly passes (HTML, CSV). This is within scope, but the agent will transmit user data (contacts, HTML, and any imported website content when 'projects import' is used) to Migma's service — which is expected behavior for an email service but worth noting for privacy.

Install Mechanism

Installation is via an npm package (@migma/cli) that creates the migma binary. Using npm for a CLI is reasonable, but npm packages can run lifecycle scripts and execute code on install; verify package provenance, version, and integrity (npm registry page, package maintainers, checksums) before installing in sensitive environments.

Credentials

Only one environment variable (MIGMA_API_KEY) is required and listed as the primary credential — appropriate for a remote email service. Remember this key grants the ability to send emails and manage contacts/domains via the service, so treat it as highly sensitive and prefer scoped/service-account keys if available.

Persistence & Privilege

The skill does not request always:true, no system config paths are required, and it doesn’t attempt to modify other skills or global agent settings. It is user-invocable and can be run autonomously by the agent (default) — normal for skills, but be mindful of autonomous runs if you don't want the agent sending emails without approval.

Assessment

This skill appears to do what it says: it calls a migma CLI and uses a single API key. Before installing: 1) verify the npm package @migma/cli on the official npm/GitHub page and confirm the maintainer and version; 2) treat MIGMA_API_KEY as a high-privilege secret (it can send email and modify contact lists); preferably use a scoped or test API key, not a production key; 3) be aware that importing CSVs or using 'projects import' will upload local data and website assets to the service — review Migma's privacy/terms; 4) consider disabling autonomous invocation for this skill if you don’t want the agent to send emails or import contacts without explicit user approval. Small note: registry metadata showed no homepage while the SKILL.md lists https://migma.ai — that mismatch is not necessarily malicious but you may want to confirm the official project URL before proceeding.

Create and send professional, on-brand emails with AI. Your agent can design emails from a prompt, send them instantly through a managed domain, and manage an entire audience — all from the terminal.

Always pass --json for structured output.

First-time setup

If the user hasn't set up yet, run these steps once:

# 1. Create an instant sending domain (no DNS needed)
migma domains managed create <companyname> --json
# → Sends from: hello@<companyname>.migma.email

# 2. Set a default project (brand)
migma projects list --json
migma projects use <projectId>

Create an email

When the user asks to create, design, or generate an email:

migma generate "Welcome email for new subscribers" --wait --json

The --wait flag blocks until the AI finishes. The JSON response includes conversationId, subject, and html.

To save the HTML locally, add --save ./email.html. To include a reference image (screenshot, design mockup), add --image <url>.

Send an email

When the user asks to send an email to someone:

# Send a generated email directly
migma send --to [e-mail] --subject "Welcome!" \
  --from-conversation <conversationId> \
  --from [e-mail] --from-name "Company" --json

# Or send from a local HTML file
migma send --to [e-mail] --subject "Hello" \
  --html ./email.html \
  --from [e-mail] --from-name "Company" --json

# Send to an entire segment or tag
migma send --segment <id> --subject "Big News" --html ./email.html \
  --from [e-mail] --from-name "Company" --json

# Personalize with template variables
migma send --to [e-mail] --subject "Hi {{name}}" --html ./email.html \
  --from [e-mail] --from-name "Company" \
  --var name=Sarah --var discount=20 --json

--from-conversation auto-exports the HTML from a generated email — no separate export step.

Validate an email

When the user wants to check an email before sending:

migma validate all --html ./email.html --json
migma validate all --conversation <conversationId> --json

Returns an overall score plus individual checks: compatibility (30+ email clients), broken links, spelling/grammar, and deliverability/spam score. Individual checks: migma validate compatibility, links, spelling, deliverability.

Export to platforms

When the user wants to export to an ESP or download a file:

migma export html <conversationId> --output ./email.html
migma export klaviyo <conversationId> --json
migma export mailchimp <conversationId> --json
migma export hubspot <conversationId> --json
migma export pdf <conversationId> --json
migma export mjml <conversationId> --json

Manage contacts

migma contacts add --email [e-mail] --firstName John --json
migma contacts list --json
migma contacts import ./contacts.csv --json
migma contacts remove <id> --json

Manage tags and segments

migma tags create --name "VIP" --json
migma tags list --json
migma segments create --name "Active Users" --description "..." --json
migma segments list --json

Import a brand

When the user wants to set up a new brand from their website:

migma projects import https://yourbrand.com --wait --json
migma projects use <projectId>

This fetches logos, colors, fonts, and brand voice automatically.

Error handling

On error, --json returns:

{"error": {"message": "Not found", "code": "not_found", "statusCode": 404}}