Task: /quality-stats command

Description

Implement the /quality-stats slash command that renders a formatted quality dashboard in the TUI. Supports a default summary view, drill-down per session, drill-down per model, and JSON/CSV export.

Acceptance Criteria

• Command registered in packages/opencode/src/command/index.ts as built-in command quality-stats
• Default summary view displays:
  • Overall quality score with trend indicator (▲/▼) vs. last week
  • Per-model comparison table: model name, avg score, response count, trend
  • Top 5 recent regressions with severity icon, time ago, component, description
  • Current session quality vs. project average
• Session drill-down (/quality-stats --session <id>):
  • Lists each response in session with composite score, sub-score breakdown
  • Highlights low-scoring responses with diagnostic notes
  • Shows score timeline within session
• Model drill-down (/quality-stats --model <id>):
  • Score distribution (min, max, avg, median, p25, p75)
  • Trend over time (last 7d, 30d)
  • Comparison against other models used in same project
• Export (/quality-stats --export json or --export csv):
  • Outputs all response data in machine-readable format
  • JSON: array of response objects with scores and metadata
  • CSV: one row per response with all fields
• Output is nicely formatted for terminal (box drawing, alignment, color via existing TUI patterns)
• Handles empty database gracefully: "No quality data yet. Quality monitoring will start tracking on your next session."

Technical Details

• Register as a built-in command in the command registry (follow pattern of init and review commands)
• Parse arguments for --session, --model, --export flags
• Query SQLite store for all data — use aggregation queries from store.ts
• Format output as markdown (the TUI renders markdown)
• For trend calculation: compare current 7-day avg vs previous 7-day avg
• For "time ago" display: use relative time formatting (e.g., "2h ago", "1d ago")

Dependencies

• Task 001: SQLite storage layer (data source)
• Task 002: Quality configuration (settings)
• Task 003: Scoring engine (score data)
• Task 005: Regression detection (regression data)

Effort Estimate

• Size: L
• Hours: 8-10
• Parallel: false (depends on 001-003, 005)