refactor: Update global-report.sh with RFC #41 capabilities

[evansenter]

Summary

• Add agent activity section showing subagent vs main session tokens
• Add trends section comparing to previous period
• Add session classification showing debugging/development/mixed
• Add failure patterns section with error rates and rework detection
• Remove languages (curiosity only), sessions (verbose), mcp-usage (secondary)
• Fix formatter predicate collision in cli.py

Test plan

• All 289 tests pass
• Ran ./scripts/global-report.sh and verified output

🤖 Generated with Claude Code

[claude]

Prompt: evansenter/dotfiles/.../claude-review.md

Code Review

Summary

This PR refactors global-report.sh to leverage RFC #41 capabilities by adding agent activity, trends, session classification, and failure pattern sections while removing less actionable sections (languages, verbose sessions, mcp-usage). It also fixes a formatter predicate collision in cli.py.

Issues Found

Critical

None

Important

None

Suggestions

• scripts/global-report.sh:96 - The file activity limit changed from 20 to 15 without documented rationale. The original limit may have been chosen deliberately. Consider documenting why 15 was chosen or keeping 20 for consistency.

• src/session_analytics/guide.md:267 - Existing documentation inconsistency (not introduced by this PR): The guide shows summary: {debugging: 5, ...} but the actual classify_sessions function returns category_distribution as the key. Consider addressing this in a follow-up to keep docs accurate.

Verdict

REQUEST_CHANGES - Minor suggestions for documentation and consistency.

---

Automated review by Claude Code