---

name: gtm-copilot description: A comprehensive skill for developing, managing, importing, and exporting Google Tag Manager configurations - GTM Copilot license: MIT compatibility: "python >= 3.x" metadata: version: 0.2.0 author: SEM Technology homepage: https://github.com/sem-technology/gtm-copilot disable-model-invocation: false allowed-tools:

• run_command
• list_dir
• view_file
• grep_search
• codebase_search

---

GTM Copilot Skill

This skill supports the entire Google Tag Manager (GTM) development workflow. All necessary tools and documentation are bundled within this skill directory.

Directory Structure (Bundled Resources)

This skill is configured as a self-contained package. Agents must always use relative paths within this skill directory.

• SKILL.md: Skill definition (this file)
• .env.example: Template for environment variable settings
• LICENSE.txt: License information
• scripts/: Folder containing all program code
  • bin/: Executable scripts (auth.py, export.py, import.py)
  • gtm_client.py: Core implementation of the GTM API client
  • authentication.py: Authentication module
  • helpers/: Utilities and client logic
• resources/: Folder for supplemental documents and sample data
  • documents/: Documents for GTM Copilot
    • audit_checkpoints.md: Detailed audit criteria for GTM configurations
  • server_container/: Configuration for Server-side GTM
    • tags.json
    • triggers.json
    • variables.json
    • built_in_variables.json
  • web_container/: Configuration for Web GTM
    • tags.json
    • triggers.json
    • variables.json
    • built_in_variables.json

Configuration

Environment variables are loaded from the .env file located in the project root.

• GTM_CLIENT_ID: OAuth 2.0 Client ID for GTM API access.
• GTM_CLIENT_SECRET: OAuth 2.0 Client Secret for GTM API access.
• GTM_REFRESH_TOKEN: OAuth 2.0 Refresh Token for automatic token renewal.
• GTM_EXPORT_ROOT_PATH: Environment variable to specify the root directory for GTM files.
  • Placeholder: Supports [[GTM_ID]] for dynamic path resolution based on the container ID.
  • Resolution Priority: CLI Argument > Env Var > Default (tmp/[[GTM_ID]]).

Command Details

1. auth (Authentication Setup)

Performs OAuth2 authentication to access the GTM API.

• Execution: python ./scripts/bin/auth.py
• Role: Input Client ID/Secret and generate a refresh token.

2. export (State Extraction)

Exports existing GTM configurations as local JSON files.

• Execution: python ./scripts/bin/export.py --url <GTM_WORKSPACE_URL>
• Output: The path defined by --output or GTM_EXPORT_ROOT_PATH (defaults to ./tmp/GTM-XXXXXX/).
• Role: Save the current state of tags, triggers, and variables as a snapshot for editing.

3. import (Change Synchronization)

Updates the GTM container based on local JSON files.

• Execution: python ./scripts/bin/import.py --url <GTM_WORKSPACE_URL>
• Input: The path defined by --directory or GTM_EXPORT_ROOT_PATH (defaults to ./tmp/GTM-XXXXXX/).
• Role:
  • Create new components
  • Update existing components (only if changes detected)
  • Automatically resolve ID references (from name-based to numeric IDs)

Workflow

Development Workflow

1. Export: Run scripts/bin/export.py to get the latest GTM state.
2. Implementation (AI Work): Modify the JSON files under tmp/.
   • Add new tags
   • Change trigger conditions
   • Fix variable definitions
3. Import: Run scripts/bin/import.py to synchronize changes.
   • After sync, official IDs and fingerprint values are written back to local JSONs.

Audit Workflow

1. Export: Run scripts/bin/export.py to get the latest GTM state.
2. Audit:
   • The agent uses the view_file tool to read resources/documents/audit_checkpoints.md to understand the audit criteria.
   • The agent then analyzes the exported JSON files in tmp/ based on those criteria.
3. Report: The agent provides a summary of findings (issues, risks, and recommendations) to the user.

Crucial: The agent must not modify any tags, triggers, or variables, nor perform any import operations, unless specifically instructed to do so by the user.

Notes on JSON File Generation & Editing

• Name-based References (Recommended):
  • When specifying firing triggers for tags, you can use the "Trigger Name" instead of a numeric ID. This is especially useful for tags referencing newly created triggers that don't have IDs yet.
  • Use the {{Variable Name}} format to reference variables.
• Dependency Resolution:
  • If a new tag references a new trigger, the import script automatically detects the dependency and creates the trigger first.
• Change Detection:
  • The script compares content hashes and skips API calls if there are no actual changes.
• Implementation References:
  • Use the JSON files in the resources/server_container/ or resources/web_container/ directories as references for correct schemas and parameter structures.
  • Important: Available tags, triggers, and variables differ significantly between Server and Web containers. Ensure you reference the correct directory for your specific container type when implementing new components.