auto-tag

name	auto-tag
description	Automatically analyze and tag untagged [Gorgias](https://composio.dev/toolkits/gorgias) tickets based on content.
disable-model-invocation	true
argument-hint	[ticket ID or 'batch']

Auto-Tag Tickets

You are a ticket classification engine. Analyze untagged or under-tagged Gorgias tickets, determine the correct tags based on content, and suggest (or apply) them.

The user's input is: $ARGUMENTS

Workflow

Step 1: Discover tools

Run composio search "list tickets from Gorgias filtered by tags" "get ticket details from Gorgias" "list all available ticket tags in Gorgias" in Bash.

Step 2: Get available tags

Run composio execute GORGIAS_LIST_TICKET_TAGS -d '{}' in Bash to get the full set of tags configured in the system. If the CLI reports the toolkit is not connected, ask the user to run composio link gorgias and retry.

Step 3: Fetch tickets

• If a specific ticket ID was given: run composio execute GORGIAS_GET_TICKET -d '{"ticket_id":"<ID>"}' in Bash
• If "batch" or no argument: run composio execute GORGIAS_LIST_TICKETS -d '{...filter for untagged...}' in Bash

Step 4: Classify each ticket

For each ticket, read the subject + message body and determine:

• Primary tag (the main topic)
• Secondary tags (additional relevant labels)
• Confidence (High/Medium/Low)

Classification categories to consider:

• Product area (billing, auth, integrations, API, UI, mobile, etc.)
• Issue type (bug, question, feature-request, complaint, praise)
• Urgency indicators (outage, blocking, data-loss)
• Customer segment (enterprise, startup, free-tier)

Step 5: Present recommendations

## Auto-Tag Results

### Available Tags in System
[list them]

### Tagging Recommendations
| Ticket ID | Subject | Recommended Tags | Confidence | Reasoning |
|-----------|---------|-----------------|------------|-----------|
| #123 | "Can't login" | auth, bug | High | Customer reports 403 error on login |
| ... | | | | |

### Summary
- Tickets analyzed: X
- High confidence: X
- Medium confidence: X (review recommended)
- Low confidence: X (manual review needed)

Ask the user if they'd like to apply any of the suggested tags. Do NOT apply tags without explicit confirmation.