weekly-ops-review

name	weekly-ops-review
description	Produce the Monday-morning support operations review — pulls the last 7–30 days of cases, correlates them against internal Teams chatter, and generates the standard artifact bundle (analysis, CSV, chart, recommendations, exec summary). Trigger on asks like "weekly ops review", "Monday support review", "prep the ops readout", or any request for a consolidated multi-section support status report.

Weekly Support Ops Review

Use this skill when the user asks for the weekly / Monday support operations review, or for a consolidated support status report covering cases, internal context, and forecast.

The goal is a leadership-ready readout: what's happening, why, who owns it, and when it clears.

Required inputs

Default to these if the user doesn't specify otherwise:

• Lookback window: 30 days (ask if user wants 7 or 14)
• Severity focus: P1 + P2 highlighted separately, P3/P4 rolled up
• Forecast horizon: 7 days

Workflow

Run these steps in order. Each step depends on the previous one's output.

1. Pull the case data

Call get_support_cases with days=30 (or the requested window). Capture:

• Total case count and the WoW delta
• Top 3–5 categories by volume
• P1/P2 count and which customers are affected
• Any case with status still open or escalated after 48h

2. Correlate with internal context

For each of the top categories identified in step 1, call query_workiq with a query naming that category (e.g. "authentication", "api timeout", "billing"). Pull out:

• Who owns the underlying issue (team lead, SRE)
• Stated root cause
• ETA or mitigation already in flight

Match each support spike to an engineering signal. A spike without a matching internal conversation is itself a finding — flag it.

3. Generate the artifact bundle

Call generate_trend_forecast with:

• case_summary: 2–3 sentence synthesis combining steps 1 and 2
• top_categories: ordered list, highest-severity first (not just highest volume)
• trend_data: JSON string of daily counts by category for the lookback window

This writes five files into workspace/:

• support-analysis.md — full analysis report
• case-data.csv — daily trend data
• weekly-forecast.png — 7-day forecast chart
• recommendations.md — action items with owners
• executive-summary.md — leadership one-pager

4. Summarize for the user

In the chat reply, give the user:

1. Bottom line — one sentence: status color (green / yellow / red) + the dominant driver
2. Top 2 incidents — each with owner, ETA, and customer impact
3. Forecast call — will volume recover this week, or slip?
4. Artifacts written — list the five files with paths so the user knows where the deep detail lives

Keep this summary tight (under ~200 words). The artifacts carry the detail.

Style

• Lead with the conclusion, not the methodology. The user knows how the sausage is made.
• Name owners by role + name when known (e.g. "Sarah Chen (Auth Lead)"). Anonymous ownership is a red flag worth surfacing.
• Numbers need context: "52 auth cases in 14 days" is noise; "52 cases, 3× baseline" is a signal.
• Never inflate severity to grab attention. A quiet week is a fine report — say so.

When NOT to use this skill

• Single-case triage or "what's the status of SC-10063" — use get_support_cases directly with filters.
• Deep-dive on one category — call query_workiq directly; you don't need the full bundle.
• Ad-hoc customer-specific reports — this skill is cross-cutting and customer-agnostic.