Skip to content

v0.1.0 — Evaluation closed loop + public onboarding

Latest

Choose a tag to compare

@WL2O2O WL2O2O released this 12 Nov 10:55
· 10 commits to main since this release

First public release. QASkillKit = QA skill toolkit for AI agents, with a TAPD-free evaluation entry anyone can run.

Highlights

  • Dual-channel test-case evaluator: 8 rule assertions (zero keys) + LLM-as-Judge (4 dimensions, key-gated; first batch human-calibrated 8/8)
  • Three-layer quality gate: local pre-commit hook + GitHub Actions CI (registered evals + whole-repo structural scan) + local judge with your key
  • 30-second trial: pip install pyyaml && python3 evals/evaluate.py — no TAPD, no key, no IDE
  • Bring your own cases: --add one-command registration; --scan-root for any directory
  • 11 agent skills: Qoder + Claude Code (committed symlink); English eval skill with bilingual triggers
  • Privacy by design: private config gitignored; keys read from env only

Honest boundaries

  • Judge scores are reference-only until calibrated on your own samples
  • TAPD-dependent skills need your own TAPD + MCP setup; the evaluator does not