Ethan has ~600 custom tools and several meta-tools that measure them separately (tool-half-life, tool-autopsy, tool-dupe-radar, tool-graveyard-reaper, dormant-skill-resurfacer) — but nothing correlates a tool's death with the git commit or session that killed it. This weekly job joins tool last-invocation timestamps against ~/.claude session logs and git history to answer 'what did you build that replaced this?', turning silent tool rot into a one-line causal story per dead tool.
Create ~/.local/bin/tool-succession-map (bash+python, ~90 lines). For each executable in ~/.local/bin, get last-run mtime from the atime/exec ledger already used by tool-half-life (reuse its cache file rather than re-scanning). For tools idle >30d, grep ~/.claude/projects/*/**.jsonl for the tool's name in the 14 days preceding its last use, and `git -C ~/.claude log --since` for commits in that window. Emit ~/Tablet/briefs/tool-succession-YYYY-WW.md with rows: dead-tool | last used | co-occurring session topic | candidate successor tool (fuzzy name/purpose match via existing tool-dupe-radar scoring). Wire a systemd timer (weekly, Sun 07:00) modeled on tool-half-life.timer, and pipe the markdown through morning-tablet-brief so it lands on the NXTPAPER for stylus triage — checkmark = reap, X = revive, routed back via stylus-reply-router into tool-graveyard-reaper.