fix(metrics): propagate --days and --project filters to all analyzers

- Pre-filter sessions by days before passing to AnalyzeEfficiency,
  AnalyzeCommits, AnalyzeConfidence, AnalyzeOutcomes, etc. Previously
  only AnalyzeVelocity applied the days window; all other sections
  silently reported all-time data regardless of --days.
- Filter facets and agentTasks by session-ID sets derived from the
  filtered session slice, so Satisfaction, Friction Trends, and Agent
  Performance also respect --days and --project.
- Export FilterSessionsByDays from the analyzer package.
- Truncate tool names longer than 22 chars to prevent mid-name line
  wrapping in session inspect and metrics efficiency sections.
- Add priority_label string to suggest --json output alongside the
  raw priority integer.
- Update fix --ai default model from claude-sonnet-4-20250514 to
  claude-sonnet-4-6.

Author: blackwell-systems

CHANGELOG.md

@@ -2,6 +2,18 @@
 
 All notable changes to claudewatch are documented here.
 
+## [Unreleased]
+
+### Fixed
+
+- **`claudewatch metrics --days` filter scope** — the `--days` window was applied only inside `AnalyzeVelocity` (Session Volume and Productivity sections) but not to the session slice passed to any other analyzer. Efficiency, Satisfaction, Token Usage, Commits, Confidence, Friction Trends, Cost per Outcome, and CLAUDE.md Effectiveness all silently reported all-time data regardless of `--days`. Fix: sessions are now pre-filtered by days immediately after the project filter, before any analyzer is called. `facets` and `agentTasks` are filtered to the same window via session-ID sets. Also fixes `--project` not filtering Agent Performance — `agentTasks` was loaded globally without session correlation.
+
+- **Long MCP tool names wrap mid-name in session inspect and metrics efficiency** — `StyleLabel` has a fixed `Width(24)`, so tool names longer than 24 characters (e.g. `mcp__commitmux__commitmux_search`) would wrap partway through the name onto the next line. Fix: names longer than 22 characters are now truncated to 22 + `..` before rendering, keeping output within the label column.
+
+- **`suggest --json` missing human-readable priority label** — the JSON output serialized `Priority` as a raw integer (`2`, `3`, `4`) with no string equivalent, while the text output showed `[HIGH]`, `[MEDIUM]`, `[LOW]`. Fix: JSON output now includes a `priority_label` field (`"HIGH"`, `"MEDIUM"`, `"LOW"`, `"CRITICAL"`) alongside the integer.
+
+- **`fix --ai` default model was stale** — default model was `claude-sonnet-4-20250514`; updated to `claude-sonnet-4-6`.
+
 ## [0.7.3] - 2026-03-02
 
 ### Fixed

internal/analyzer/velocity.go

@@ -9,7 +9,7 @@ import (
 // AnalyzeVelocity computes productivity metrics from session data, filtered
 // to the last N days. If days is 0 or negative, all sessions are included.
 func AnalyzeVelocity(sessions []claude.SessionMeta, days int) VelocityMetrics {
-	filtered := filterSessionsByDays(sessions, days)
+	filtered := FilterSessionsByDays(sessions, days)
 
 	metrics := VelocityMetrics{
 		TotalSessions: len(filtered),
@@ -39,9 +39,9 @@ func AnalyzeVelocity(sessions []claude.SessionMeta, days int) VelocityMetrics {
 	return metrics
 }
 
-// filterSessionsByDays returns sessions whose StartTime falls within the last
+// FilterSessionsByDays returns sessions whose StartTime falls within the last
 // N days. If days <= 0, all sessions are returned.
-func filterSessionsByDays(sessions []claude.SessionMeta, days int) []claude.SessionMeta {
+func FilterSessionsByDays(sessions []claude.SessionMeta, days int) []claude.SessionMeta {
 	if days <= 0 {
 		return sessions
 	}

internal/app/fix.go

@@ -42,7 +42,7 @@ func init() {
 	fixCmd.Flags().BoolVar(&fixFlagAll, "all", false, "Fix all projects with score < 50")
 	fixCmd.Flags().BoolVar(&fixFlagJSON, "json", false, "Output proposed changes as JSON")
 	fixCmd.Flags().BoolVar(&fixFlagAI, "ai", false, "Use Claude API for project-specific CLAUDE.md generation")
-	fixCmd.Flags().StringVar(&fixFlagModel, "model", "claude-sonnet-4-20250514", "Claude model to use for AI generation")
+	fixCmd.Flags().StringVar(&fixFlagModel, "model", "claude-sonnet-4-6", "Claude model to use for AI generation")
 	rootCmd.AddCommand(fixCmd)
 }
 

internal/app/metrics.go

@@ -88,6 +88,9 @@ func runMetrics(cmd *cobra.Command, args []string) error {
 		sessions = filterSessionsByProject(sessions, metricsProject)
 	}
 
+	// Filter by days — applied early so all downstream analyzers see the same window.
+	sessions = analyzer.FilterSessionsByDays(sessions, metricsDays)
+
 	// Load facets.
 	facets, err := claude.ParseAllFacets(cfg.ClaudeHome)
 	if err != nil {
@@ -98,15 +101,22 @@ func runMetrics(cmd *cobra.Command, args []string) error {
 		facets = scanner.FilterFacetsByProject(facets, sessions, metricsProject)
 	}
 
+	// Filter facets to the same session window as the day-filtered sessions.
+	facets = filterFacetsBySessionIDs(facets, sessions)
+
 	// Load agent tasks from session transcripts.
 	agentTasks, err := claude.ParseAgentTasks(cfg.ClaudeHome)
 	if err != nil {
 		// Non-fatal if transcript parsing fails.
 		agentTasks = nil
 	}
 
+	// Filter agent tasks to the active session window.
+	agentTasks = filterAgentTasksBySessionIDs(agentTasks, sessions)
+
 	// Run analyzers.
-	velocity := analyzer.AnalyzeVelocity(sessions, metricsDays)
+	// Sessions are pre-filtered by days above; pass 0 to skip the internal re-filter.
+	velocity := analyzer.AnalyzeVelocity(sessions, 0)
 	efficiency := analyzer.AnalyzeEfficiency(sessions)
 	satisfaction := analyzer.AnalyzeSatisfaction(facets)
 	agents := analyzer.AnalyzeAgents(agentTasks)
@@ -254,8 +264,12 @@ func renderEfficiency(e analyzer.EfficiencyMetrics) {
 			limit = len(sorted)
 		}
 		for _, kv := range sorted[:limit] {
+			name := kv.key
+			if len(name) > 22 {
+				name = name[:22] + ".."
+			}
 			fmt.Printf("   %s %s\n",
-				output.StyleLabel.Render(kv.key),
+				output.StyleLabel.Render(name),
 				output.StyleValue.Render(fmt.Sprintf("%d", kv.value)))
 		}
 	}
@@ -796,6 +810,42 @@ func computeTokenUsage(sessions []claude.SessionMeta) tokenUsage {
 	}
 }
 
+// filterFacetsBySessionIDs keeps only facets whose SessionID is in the given sessions.
+func filterFacetsBySessionIDs(facets []claude.SessionFacet, sessions []claude.SessionMeta) []claude.SessionFacet {
+	if len(sessions) == 0 {
+		return nil
+	}
+	ids := make(map[string]struct{}, len(sessions))
+	for _, s := range sessions {
+		ids[s.SessionID] = struct{}{}
+	}
+	var filtered []claude.SessionFacet
+	for _, f := range facets {
+		if _, ok := ids[f.SessionID]; ok {
+			filtered = append(filtered, f)
+		}
+	}
+	return filtered
+}
+
+// filterAgentTasksBySessionIDs keeps only agent tasks whose SessionID is in the given sessions.
+func filterAgentTasksBySessionIDs(tasks []claude.AgentTask, sessions []claude.SessionMeta) []claude.AgentTask {
+	if len(sessions) == 0 {
+		return nil
+	}
+	ids := make(map[string]struct{}, len(sessions))
+	for _, s := range sessions {
+		ids[s.SessionID] = struct{}{}
+	}
+	var filtered []claude.AgentTask
+	for _, t := range tasks {
+		if _, ok := ids[t.SessionID]; ok {
+			filtered = append(filtered, t)
+		}
+	}
+	return filtered
+}
+
 // detectClaudeMDChanges finds projects with CLAUDE.md files and returns their
 // modification times as change events for effectiveness analysis.
 func detectClaudeMDChanges(projects []scanner.Project) []analyzer.ClaudeMDChange {

internal/app/sessions.go

@@ -308,7 +308,11 @@ func renderInspect(r sessionRow) {
 			limit = len(tools)
 		}
 		for _, t := range tools[:limit] {
-			muted(t.name, fmt.Sprintf("%d", t.count))
+			name := t.name
+			if len(name) > 22 {
+				name = name[:22] + ".."
+			}
+			muted(name, fmt.Sprintf("%d", t.count))
 		}
 	}
 

internal/app/suggest.go

@@ -314,7 +314,18 @@ func filterByProject(suggestions []suggest.Suggestion, project string) []suggest
 func outputSuggestJSON(suggestions []suggest.Suggestion) error {
 	enc := json.NewEncoder(os.Stdout)
 	enc.SetIndent("", "  ")
-	return enc.Encode(suggestions)
+	type suggestionOut struct {
+		suggest.Suggestion
+		PriorityLabel string `json:"priority_label"`
+	}
+	out := make([]suggestionOut, len(suggestions))
+	for i, s := range suggestions {
+		label := priorityToLabel(s.Priority)
+		// Strip brackets from the label for clean JSON (e.g. "HIGH" not "[HIGH]").
+		label = label[1 : len(label)-1]
+		out[i] = suggestionOut{s, label}
+	}
+	return enc.Encode(out)
 }
 
 func renderSuggestions(suggestions []suggest.Suggestion) {