backlog-grooming-assistant

Backlog Grooming Assistant

This skill reads a local backlog export (CSV or Markdown table), flags problematic items using a structured checklist, and produces a grooming session agenda with a scorecard. No external service connections required — works entirely from pasted or uploaded file contents.

Input:

• Backlog file contents: CSV (with column headers) or Markdown table — pasted directly or uploaded
• Minimum required columns: item ID and title; additional columns (owner, priority, estimate, status, last-updated) enable more flags

Output:

• Markdown document with: Scorecard → Top Issues → Grooming Agenda → Skipped Checks

---

Language Detection

Detect the user's language from their message:

• If Russian (or contains Cyrillic): respond in Russian
• If English (or other Latin-script language): respond in English
• If ambiguous: respond in the language of the trigger phrase used

The detected language applies to all output text: section headings, table column headers, field labels, agenda topic titles, error messages, and recommended actions. Translate every part of the output format template into the detected language — do not leave any section header or label in English when responding in Russian.

---

Instructions

Step 1: Parse and Validate Input

1. Accept backlog content from the user (pasted text or file contents)

   • If content is empty: stop and return — "Backlog is empty — no items to analyze. Paste CSV or Markdown table content."

2. Detect format:

   • CSV: first line contains comma-separated headers
   • Markdown table: lines starting with | and separator row |---|
   • If neither format detected: stop — "Format not recognized. Supported: CSV with headers, Markdown table. Paste the file contents in chat."

3. Parse rows: extract all items as a list of key-value pairs

   • Identify available columns by name (case-insensitive matching)
   • Map common synonyms: "Task" → title, "Assignee" → owner, "SP"/"Points" → estimate, "Updated"/"Modified" → last-updated
   • If columns cannot be mapped automatically: ask user to specify column mapping for the ambiguous columns, then continue