report-and-summarize

name: report-and-summarize description: | Generates reports, summaries, and analysis from {{Service Name}} data. Use when the user asks to "summarize", "report on", "give me an overview", "analyze", "break down", "how are we doing on", "what's the trend", or "weekly/monthly update" for {{Service Name}} data. metadata: author: "{{Your Name}}" version: "1.0" pattern: aggregation

Report and Summarize {{Service Name}} Data

What This Skill Does

Transforms raw {{Service Name}} data into executive summaries, trend analysis, and structured reports. Designed for users who need to understand what's happening without querying dashboards or pulling exports themselves.

When to Activate

• User asks for a summary, overview, or status report
• User wants to know trends, patterns, or distributions
• User asks "how are we doing" on a metric or category
• User needs a weekly, monthly, or ad-hoc report
• User asks to compare periods, teams, or categories

Workflow

1. Clarify the report scope. Determine:

   • What entity: tickets, deals, orders, customers, etc.
   • Time range: last week, this month, quarter-to-date, custom range
   • Grouping: by status, assignee, category, priority, region
   • Audience: is this for the user, their manager, or a team?

   If the user says "give me a summary," default to the current week and ask if they want a different range.

2. Retrieve the data. Use the search_{{entity}} tool with date filters and relevant grouping parameters. Pull enough data to provide meaningful aggregation.

3. Analyze and aggregate. Calculate:

   • Totals and counts by category
   • Changes from the previous period (week-over-week, month-over-month)
   • Top items (highest priority, most recent, largest value)
   • Items needing attention (overdue, stalled, unassigned)

4. Present the report. Structure output for the stated audience. Lead with the headline number, then drill into details.

5. Suggest actions. Based on what the data shows, recommend concrete next steps.

Output Format

Executive Summary (Default)

{{Service Name}} Weekly Summary — Apr 28 – May 4, 2026

Headline: 47 new tickets this week (+12% vs. last week). 8 are unresolved past SLA.

Top Issues:

1. Login failures — 12 tickets, all High priority, assigned to Platform team
2. Billing discrepancies — 8 tickets, 3 past SLA
3. API timeout errors — 6 tickets, started May 1

Recommended Actions:

• Escalate the 8 past-SLA tickets — want me to reassign or notify owners?
• Login failures cluster started May 1 — may be related to the API timeouts
• Schedule a review of the billing process (recurring pattern, 3rd week in a row)

Detailed Breakdown (When Asked)

Provide per-assignee, per-category, or per-priority breakdowns as additional tables. Keep each table focused on one dimension.

Trend Report (For Period Comparisons)

Use a text-based trend format since Cowork doesn't render charts:

New tickets per week (last 4 weeks):
Apr 7:  ████████████████████████ 38
Apr 14: ██████████████████████████ 41
Apr 21: ██████████████████████████████ 42
Apr 28: ██████████████████████████████████ 47

Handling Edge Cases

• No data in range: Report it clearly. Suggest expanding the date range or checking if the filters are too narrow.
• Insufficient data for trends: If fewer than 2 periods exist, skip the comparison. Say "Not enough history for trend analysis yet."
• Large datasets: Aggregate server-side when possible. If the API doesn't support aggregation, retrieve up to 200 records and note that the report covers a sample.
• User asks for a chart or graph: Explain that Cowork displays text-based reports. Offer to format the data as a table that can be pasted into Excel, or suggest creating a document with the data.

Handling Authentication

If a tool call fails because the user hasn't connected to {{Service Name}} yet:

1. Tell the user: "I need to connect to {{Service Name}} to pull the data for this report. You should see a sign-in prompt — please complete it and I'll try again."
2. Do NOT retry the data retrieval until the user confirms sign-in.
3. If partial data was already retrieved before the auth failure, present what you have and note which sections are incomplete.