- Stage: usable early package.
- Active focus: move Vibe Harness from a template-only skill toward a small executable audit loop.
- Last verified: 2026-05-08 with
python3 scripts/validate.py, skill quick validation,python3 scripts/audit.py examples/expected-harness, andpython3 scripts/audit.py ..
- Created the platform-neutral
vibe-harnessskill package. - Added English and Chinese README files.
- Added compatibility notes for Codex, Claude Code, Cursor, Copilot, Gemini, and other file-reading coding agents.
- Published GitHub repo and releases through
v0.4.0. - Broadened positioning from vibe-coding-only to AI coding projects.
- Added a static harness audit script.
- Which real-world project should be used as the first public before/after case study?
- How strict should audit scoring become before
v1.0.0? - Should the audit stay Python-only or eventually expose a small packaged CLI?
- The audit is intentionally static and does not run project commands.
- Verification command detection only covers common package, Makefile, Python, Rust, and Go patterns.
- The project still needs real external feedback before more launch material is added.
- Run Vibe Harness on one real repo and record the before/after result.
- Improve audit precision based on real false positives.
- Share one plain launch post asking for feedback, not stars.