github alexgreensh/token-optimizer v5.3.9
v5.3.9 — Editorial polish + stale-regen actually fires

latest releases: v5.11.13, v5.11.12, v5.11.11...
2 months ago

Three fixes after walking through what a real user saw in the recommendations.

What changed

  • Broken examples/ reference gone. The file-exclusion recommendation pointed at a token-optimizer/examples/ directory that does not exist in the repo. Replaced with an inline starter JSON you can drop straight into .claude/settings.json, so there's nothing to hunt for.
  • Git-instructions recommendation reads as prose. Previously, a forced-newline markdown layout rendered as orphaned fragments like "json." and "json:" under WHY/TRADE-OFFS headings (the v5.3.7 single-Details block exposed the underlying weirdness). Rewritten as one clean paragraph.
  • Stale-regen actually rewrites the file. v5.3.8 added an ensure-health check that regenerates the dashboard when the on-disk version doesn't match, but a 60-second throttle in generate_standalone_dashboard silently returned before rewriting. Added force=True so version-mismatch regen always wins. Verified: inject a fake "v5.1.0" marker, run ensure-health, the file is now v5.3.9.

Upgrade

/plugin update token-optimizer, then close and reopen Claude Code. First SessionStart refreshes the dashboard HTML. Hard-reload the tab (Cmd-Shift-R / Ctrl-Shift-R). Verify with:

python3 $MEASURE_PY version    # -> 5.3.9

Don't miss a new token-optimizer release

NewReleases is sending notifications on new releases.