Converting email templates

Purpose

Automatically convert email templates from other email service providers into Mailtrap-compatible Handlebars syntax. Handles both deterministic pattern conversions (via a Python script) and complex edge cases (via LLM intelligence).

Related skills: using-email-templates (Handlebars syntax, API template send), sending-emails (stream bases and auth), authorizing-api-requests (tokens, env vars, account_id resolution — needed when creating templates or sending via the Templates API after conversion).

Input

• Source provider (required): sendgrid, mailgun, mandrill-handlebars, mandrill-mergetags, postmark, brevo, or ses
• Input: a single template file path, a directory of .html template files, or raw HTML pasted inline
• Output directory (optional): defaults to a /converted/ subdirectory next to the input

If the user does not specify the source provider, ask before proceeding. If the user provides template HTML inline (pasted in chat), save it to a temporary .html file first.

Workflow

Step 1: Detect mode

Fast mode (default): For providers with deterministic conversion patterns. The Python script handles these reliably.

Deep mode: Triggered when:

• Provider is brevo AND the template contains nested loops, multiple filters, or complex conditional nesting
• The script flags patterns that could benefit from LLM judgment
• The user explicitly requests deep mode
• Inline HTML is provided (run LLM conversion directly)

Step 2: Run the converter script

Execute the Python converter script:

python3 {skill_dir}/scripts/converter.py \
  --provider {provider} \
  --input {input_path} \
  --output {output_dir}

Where {skill_dir} is the directory containing this SKILL.md file.

Read the output:

• Exit code 0: clean conversion, no flags. Proceed to Step 4.
• Exit code 1: conversion completed with flags. Proceed to Step 3.
• Exit code 2: error. Report to user and stop.

Read the .report.txt file(s) generated alongside each converted template. For bulk conversions, also read _summary.txt.

Note: When flags are present, the script automatically:

• Prepends a <!-- WARNING: ... --> comment at the top of the converted file
• Wraps unsupported syntax in <!-- FLAGGED [RULE]: message --> HTML comments so the file can be pasted into the Mailtrap editor without triggering parser errors

Step 3: Handle flags (Deep mode)

For each flag in the report:

1. Read the flagged line in the original template.
2. Consult references/conversion-rules.md for the specific rule and recommended action.
3. Categorize the flag:
   • App-layer change required (e.g., formatDate, comparison helpers): Cannot be resolved in the template. Report to the user with a clear explanation of what they need to change in their application code. Suggest the replacement variable name.
   • Template-level fix possible (e.g., Brevo nested loops, Mandrill mc:edit regions): Apply the conversion using LLM judgment. Show the user the before/after and ask for confirmation before writing.
   • Ambiguous (e.g., Postmark closing tag mismatch): Show the user both the original context and the converted output. Ask them to verify.

Present flags grouped by category:

## Conversion Complete

### Auto-converted: [count] changes
[Brief summary or collapsed details]

### Needs your application code: [count] flags
[For each: what to change, why, suggested variable name]

### Needs verification: [count] flags
[For each: show before/after, ask for confirmation]

Step 4: Present results

Single file:

Converted: {filename}
Changes: [count] | Flags: [count]
Output: {output_path}

[Show the conversion report summary]
[If flags exist, show the categorized flag breakdown from Step 3]

Bulk conversion:

Converted: [count] files
Total changes: [count] | Total flags: [count]
Clean: [count] files | Needs review: [count] files
Output: {output_dir}

[Show the summary report]
[List files needing review with their specific flags]

Step 5: API field mapping reminder

After template conversion, always remind the user about the API-side changes they need to make. Reference the quick mapping table:

Only show the row for the provider that was just converted.

Inline conversion (no file)

If the user pastes template HTML directly in chat instead of providing a file path:

1. Identify the source provider (ask if not stated).
2. Apply the conversion rules from references/conversion-rules.md directly using LLM intelligence.
3. Comment out unsupported syntax in the converted output by wrapping it in <!-- FLAGGED [RULE]: message --> HTML comments (same format as the script). This prevents editor errors when the user pastes the output into Mailtrap.
4. If there are flags, prepend a <!-- WARNING: ... --> comment at the top of the converted output noting how many flagged patterns need resolution.
5. Present the converted Handlebars output with a list of changes made.
6. Flag any patterns that need app-layer changes.
7. Offer to save the result to a file.

This mode always uses Deep mode (LLM) since there is no file to pass to the script.

Rules

• Never silently skip a flagged pattern. Every flag must be reported to the user.
• Never modify the original file. Converted output always goes to the output directory.
• Preserve HTML structure. Only modify template syntax (variables, conditionals, loops, helpers). Do not touch HTML tags, attributes, CSS, or non-template content.
• When in doubt, flag. If a pattern is ambiguous, flag it for manual review rather than making an incorrect conversion.
• Report counts. Always show the number of changes and flags so the user knows the scope.
• One provider at a time. If a directory contains templates from multiple providers, ask the user to separate them or specify the provider for the batch.