Mark PreCompact transcript backup hook as done (TODO item 8)

Hook implemented at ~/.claude/hooks/pre-compact-backup.sh, configured
in ~/.claude/settings.json for both auto and manual compaction triggers.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

TODO.md

@@ -37,9 +37,9 @@ See: [skills-and-progressive-disclosure.md](improvements/skills-and-progressive-
 `boostvolt/claude-code-lsps` gives Claude real-time types, go-to-definition, and find-references. "The single biggest productivity gain" per community consensus.
 See: [plugins-and-language-servers.md](improvements/plugins-and-language-servers.md)
 
-## 8. Add PreCompact transcript backup hook
-**Impact: Medium | Effort: Low**
-Save conversation transcript before auto-compaction. Pairs with auto-memory to ensure no context is lost during long sessions.
+## 8. ~~Add PreCompact transcript backup hook~~ → DONE
+**Impact: Medium | Effort: Low** ✅
+Implemented in `~/.claude/hooks/pre-compact-backup.sh`. Fires on both auto and manual compaction. Reads `transcript_path` from stdin JSON, copies to `~/.claude/transcript-backups/` with timestamp, prunes backups older than 30 days.
 See: [hooks-and-automation.md](improvements/hooks-and-automation.md)
 
 ## 9. Prune and restructure project CLAUDE.md
@@ -59,4 +59,5 @@ See: [planning-and-workflow.md](improvements/planning-and-workflow.md)
 - **Best-practices library** — 11 topic files + INDEX.md extracted from cluster-bootstrap and custom-claude-skills. Referenced in CLAUDE.md for auto-loading per project.
 - **Knowledge distillation pipeline** — Three new skills (`/log`, `/reflect-logs`, `/distill-best-practices`) with settings.yaml, state tracking, log pruning, and staleness detection.
 - **Parent CLAUDE.md symlink** — `~/dev/claude/CLAUDE.md` → claude-foundations, all projects inherit automatically.
+- **PreCompact transcript backup hook** — `~/.claude/hooks/pre-compact-backup.sh` with auto+manual matchers.
 - **claude-foundations Gitea repo** — Created under skynet org, pushed.