Releases: talkstream/better-model
Releases · talkstream/better-model
v0.4.0 — Auto git-add
What's new
Auto git-add
init and audit --fix now automatically stage all created/modified files in git.
Solves a real pattern observed during field testing: .claude/ model configs and docs/BETTER-MODEL.md were not committed across 3 consecutive commits in a live Claude Code session — not because they were gitignored, but because the session didn't perceive them as deliverables.
Now all better-model files are staged automatically. Nothing else is touched.
Quick start
npx better-model@0.4.0 init51 tests, 0 dependencies, Snyk clean.
v0.3.1 — Stop waiting for Opus on every grep
What's new
Enforcement mode (v0.3.0)
npx better-model initnow injectsmodel:andeffort:frontmatter into.claude/agents/and.claude/skills/- Inference engine maps agent names to tiers: review → Sonnet, migrate → Opus, scan → Haiku
--softflag for reference-only mode (previous behavior)audit --fixfor post-install enforcement
Polished README & GitHub infrastructure (v0.3.1)
- New tagline: "Stop waiting for Opus on every grep."
- Badges: npm version, zero dependencies, install size, license
- Speed-first structure — install command within 5 lines
- GitHub Actions CI (Node 18/20/22)
- Issue templates: bug report + matrix update (evidence required)
- CHANGELOG, SECURITY policy, PR template
Field test result
A Claude Code session read the decision matrix in soft mode and proactively updated all 8 agent/skill configs on its own — no audit --fix needed. The matrix works as both documentation and instruction.
Quick start
npx better-model initFull changelog
See CHANGELOG.md