Review and audit findings are numbered now, so you can answer "fix 2 and 4" instead of quoting lines back. CodeBuddy runs ponytail through the existing Claude plugin, and /ponytail-gain shows the honest agentic benchmark averages (54% less code, 22% fewer tokens, 20% lower cost, 27% faster, safety kept) instead of the old single-shot numbers a chatty baseline inflated. The README got slimmer too: every per-agent install step now lives in INSTALL.md.
What's Changed
- docs: add Qwen Code support by @Dattuog in #888
- feat: CodeBuddy support via the existing Claude plugin (#854) by @ShantanuTaro in #891
- feat: add Goose adapter docs and schema guard (#880) by @nolanchic in #882
- docs: add Muse Code CLI support by @dangtrungan in #950
- docs: note that /ponytail-review takes a target in plain words by @DietrichGebert in #1023
- feat: number review and audit findings (#327) by @yashrao2607 in #523
- docs: new skills come from the maintainer, ideas go in an issue by @DietrichGebert in #1024
- docs: slim the README, move per-agent install steps to INSTALL.md by @DietrichGebert in #1026
- docs: name the only official install sources in the README by @DietrichGebert in #1027
- test: skip the /bin/sh hook checks where there is no /bin/sh by @DietrichGebert in #1028
- fix(gain): show the agentic benchmark averages, not the old single-shot medians by @DietrichGebert in #1029
- fix(bench): report the real error from correctness checks by @DietrichGebert in #1030
- chore: release v4.12.0 by @DietrichGebert in #1031
New Contributors
Full Changelog: v4.11.0...v4.12.0