tf-module-updater

name	tf-module-updater
description	Fetches existing Terraform modules from GitHub via MCP tools and applies targeted, surgical edits. Reads module-index.md to locate modules in the repository, uses get_file_contents to fetch current files with SHAs, and applies changes via edit_file without full file rewrites. Use when asked to update, modify, patch, or change an existing Terraform module that is already committed to GitHub. Not for new module creation.
compatibility	Requires GitHub MCP server tools (list_directory_contents, get_file_contents, create_or_update_file). Requires read_file, write_file, and edit_file tools.
metadata	{"author":"talkops-ai","version":"1.0","generated-by":"manual-authoring"}
allowed-tools	read_file write_file edit_file ls grep

Terraform Module Updater

Overview

You fetch existing Terraform modules from GitHub and apply targeted changes. You never rewrite entire files — you make surgical, minimal edits that preserve existing formatting and conventions. This is the UPDATE workflow counterpart to the tf-generator's CREATE workflow.

Workflow

Progress:

• Step 1: Locate the module in the repository
• Step 2: Fetch existing module files from GitHub
• Step 3: Analyze the existing code structure
• Step 4: Apply targeted changes
• Step 5: Update documentation
• Step 6: Return summary

1. Locate the Module

Read the module index to find the module path:

read_file /memories/module-index.md

The index contains entries like:

| Service | Repo Path | Last Updated |
|---------|-----------|-------------|
| {service} | modules/{service}/ | 2025-01-15 |

If the module is not in the index, ask the coordinator for the repository path.

Also load organizational conventions:

read_file /memories/org-standards.md

2. Fetch Existing Files from GitHub

Use GitHub MCP tools to fetch all .tf files:

1. List the directory contents:

   list_directory_contents(repo, module_path)
   

2. For each file, fetch the content AND SHA:

   get_file_contents(repo, file_path)
   

3. Write fetched files to the local workspace:

   write_file ./workspace/terraform_modules/{service}/{filename} {content}
   

Critical: Track the SHA of every file you fetch. You will need it when committing changes back to GitHub.

3. Analyze the Existing Code

Before making changes, understand the current module:

• Resource naming conventions used
• Variable naming patterns
• Tagging strategy
• Provider version constraints
• Conditional creation patterns (count or for_each)

See references/update-patterns.md for common analysis patterns.

4. Apply Targeted Changes

Use edit_file for ALL modifications. Never rewrite an entire file.

Rules for surgical edits:

• Only touch files that need changing
• Preserve existing formatting, indentation, and conventions
• When adding new resources, follow the established naming patterns
• When adding variables, maintain alphabetical or logical grouping
• When modifying count or for_each expressions, verify all dependent references (outputs, other resources, data sources)
• When changing for_each keys, verify that outputs using values() or direct key access are still valid

5. Update Documentation

If inputs or outputs changed:

• Update the README.md inputs table
• Update the README.md outputs table
• Update any usage examples that reference changed variables

6. Return Summary

Return exactly:

Updated {N} files: [list]. Changes made: [diff summary].

Gotchas

• NEVER rewrite an entire file — make targeted, surgical edits only
• ALWAYS track file SHAs for GitHub commit operations
• If get_file_contents returns an error (404), the file doesn't exist — treat it as a new file for the commit operation
• Changing a variable's type is a BREAKING CHANGE — flag it explicitly
• Removing a variable or output is a BREAKING CHANGE — flag it explicitly
• When changing count to for_each (or vice versa), Terraform will DESTROY and RECREATE all instances unless moved blocks are added to map each old address to the new one — always recommend moved blocks
• Splat expressions ([*]) work with count but NOT directly with for_each — use values(resource)[*].attr instead