Releases: jpbaking/agentic-tests
Releases · jpbaking/agentic-tests
Release list
v3.0.0 — portable agent authoring layout
Breaking packaging update: moves canonical skills to .agents/skills, adds byte-identical Claude mirrors and thin Cline workflows, supports the current Antigravity, Codex, Claude, Cline, and Pi paths, and makes usage documentation host-neutral.
Full Changelog: v2.0.0...v3.0.0
v2.0.0 — user-global install model
Breaking: the six skills now install user-global; install.sh/install.ps1 are removed and AGENT-INSTALL.md is the only install path.
- Skills copy to per-user global skill directories (
~/.agents/skills/,~/.claude/skills/,~/.gemini/config/skills/,~/.cline/skills/); nothing is written into projects — no bridges, no.gitignoreedits. - The only per-project footprint is the generated agent tests themselves (
*.agentic.spec.*, …), which are committed — they are the safety net. - Cursor supported: discovers the
~/.agents/skills/copies natively.
Migration from v1 project-scoped installs: re-run the install prompt once per user; stale gitignored project adapters can be removed manually.
v1.0.0 — universal Agent Skills release
First tagged release.
Highlights
- Six portable Agent Skills under
skills/shared/for Codex, Claude Code, Google Antigravity, and Cline:agentic-unit-test,agentic-mutation-check,agentic-refactor,agentic-test-update, and the newly skill-ifiedagentic-coverage-reportandagentic-test-clean(previously Cline-only workflows). - Installers — tarball-based
install.sh/install.ps1copy the skills to.agents/skills/and.claude/skills/, add conditionalAGENTS.md/CLAUDE.mdpointers, and gitignore the installed copies. - AGENT-INSTALL.md — preferred, merge-aware agent-guided install (see README prompt).
- Cline
.clinerules/workflows/shortcuts retired; each skill is invocable as/agentic-*in Claude Code / Antigravity / Cline, via$mention in Codex, or implicitly by description.
The test-generation guarantees are unchanged: main code and user tests are read-only, coverage is measured on agent tests alone, and every generated test passes a flakiness gate.