Fixed
- The Claude Code footer and the Codex pane showed only
TokenPakinstead of the session line when the user's locale collates regex ranges, for exampleen_US.UTF-8on Linux. The footer's ASCII check now runs in byte order under any locale.
Changed
- The README and the documentation home page now open with what TokenPak shows for a coding-agent session: measured usage, estimated cost, burn and runway.
- Personal names are removed from source comments, schema examples and test fixtures.
- The orchestration handoff's human recipient is now registered as
operator. A handoff addressed to the previous name fails as an unknown agent. - The release metadata's documentation-revision pin now names the rewritten commit ID of the same documentation content. The documentation repository's history was rewritten to remove personal data.
Compatibility
- No API, CLI, storage or configuration change apart from the handoff recipient name above.
- TokenPak Pro 0.5.1 supports TokenPak 1.26.0 through 1.30.0. It refuses to run on 1.30.1. Keep Pro installations on 1.30.0 until Pro 0.5.2, which adds 1.30.1, is installed.
- See upgrade, rollback and release status.
Known issues
tokenpak doctorreports "legacy DB" for a newly createdmonitor.dbthat has no schema-version or budget-alert table yet. The database is not legacy; no action is needed.
Security
This release accepts the open NLTK GHSA-8mgp-746c-j5xp and Accelerate GHSA-4j2p-28q2-5m79 optional-dependency findings for this release only, following a fresh October 1 dependency reassessment. Neither advisory lists a patched version, and no verified published fix is available for either. The base install excludes both packages. SECURITY.md lists the affected extras and limitations. Both findings remain open and must be resolved or reassessed before another release.