Categorize Activity Types

Activity Type Classification

Categorize Jira issues into Red Hat's Sankey capacity allocation categories and update them via MCP Jira tools. This skill supports two modes of operation:

• Single-issue mode: Categorize one issue by key (invoked by /jira:categorize-activity-type)
• Batch mode: Categorize all unclassified issues in a project (invoked by /jira:batch-categorize-activity-types)

Both modes use identical classification logic, validation, and reporting.

Valid Activity Types

These are the only valid values. Use the exact strings:

For detailed definitions, subcategories, and edge cases, see resources/activity-type-guidance.md.

Activity Type Field

The Activity Type custom field ID is customfield_10464. This is the same across all projects on redhat.atlassian.net.

Prerequisites Check

Before starting any work, verify MCP Jira tools are available:

1. Attempt to call searchJiraIssuesUsingJql with a simple query (e.g., jql: "project = OCM", maxResults: 1)
2. If the tool is not found or returns an MCP connection error, stop immediately and tell the user:
   • "The MCP Jira tools are not available. This workflow requires the Atlassian Rovo MCP server to be configured with access to redhat.atlassian.net."
3. Do NOT proceed to any phase if this check fails

Working Directory

Both modes write artifacts to .work/activity-type-classifier/. Create this directory before starting:

mkdir -p .work/activity-type-classifier

Workflow

Copy this checklist and track progress:

Classification Progress:
- [ ] Prerequisites: MCP Jira tools available
- [ ] Phase 1: Gather issues
- [ ] Phase 2: Categorize each issue
- [ ] Phase 3: Validate & generate report
- [ ] Phase 4: Apply updates (with approval)
- [ ] Phase 5: Iterate (batch mode only)

Phase 1: Gather Issues

Single-Issue Mode

Fetch the issue by key using getJiraIssue:

• Fields: summary,description,issuetype,labels,parent,components,priority,customfield_10464
• If the issue already has an Activity Type set (customfield_10464 is not null), inform the user and stop

Batch Mode

Parse user input for:

• Project key (required) — e.g., OCM, ARO
• Issue type (optional, default: Epic)
• Extra JQL filters (optional) — e.g., AND resolved >= "2025-01-01"

Always filter for issues without an Activity Type set. The "Activity Type" is EMPTY condition is mandatory in every query — do not ask the user whether to include it.

Construct the JQL query using this template:

project = {PROJECT} AND issuetype = {TYPE} AND "Activity Type" is EMPTY

Common additions:

• Date filter: AND resolved >= "2025-01-01"
• Open issues only: AND status != Closed

Execute searchJiraIssuesUsingJql with maxResults: 50. If more results exist, make a second call with nextPageToken to get up to 100 total. Combine both result sets.

From each issue, extract: key, summary, description (truncate to 2000 chars), labels, issuetype, status, priority, comment, and parent. Save all extracted data to .work/activity-type-classifier/issues.json.

Report the count of issues found to the user before proceeding.

Phase 2: Categorize Issues

Pre-check — Parent inheritance: Before classifying each issue, check if it has a parent issue. If the parent has an Activity Type set (customfield_10464), inherit it directly — no further classification needed. Set confidence to "High" and reasoning to "Inherited from {PARENT_KEY}". To look up a parent's Activity Type, call getJiraIssue with the parent's key and check customfield_10464. Cache parent lookups to reduce API calls — multiple children may share the same parent.

For remaining issues (no parent or parent has no Activity Type), read summary, description, labels, comments, and status. Apply the classification rules below and the detailed guidance in resources/activity-type-guidance.md.

Save classifications to .work/activity-type-classifier/classifications.json as a JSON array:

[
  {
    "key": "OCM-12345",
    "summary": "Issue title",
    "activityType": "Product / Portfolio Work",
    "confidence": "High",
    "reasoning": "New customer-facing feature for cluster provisioning"
  }
]

In single-issue mode, this array contains exactly one entry. In batch mode, it contains all classified issues.

If total issues exceed 50 (batch mode), process in sub-batches of 20 to manage context.

Phase 3: Validate and Report

Run the validation and report generation scripts. These are located in scripts/ relative to this skill's directory.

1. Run the validation script:

   bash plugins/jira/skills/categorize-activity-types/scripts/validate-classifications.sh .work/activity-type-classifier/classifications.json

2. Fix any validation errors before proceeding
3. Generate the report:

   python3 plugins/jira/skills/categorize-activity-types/scripts/generate-report.py .work/activity-type-classifier/classifications.json .work/activity-type-classifier/report.md

4. Display the report summary to the user