mirror of
https://github.com/pproenca/dot-skills.git
synced 2026-09-14 15:58:12 +08:00
pproenca
3efae1bf52
Standalone, calibrated-proxy quality system grounded in Anthropic's Agent Skills guidance, the dev-skill discipline rubrics, and skills-ref. - scripts/quality/: deterministic SQS scorer (score.mjs lib + score-skill, score-all, score-history, report, track CLIs) - quality/: METRICS.md spec, snapshot.json (HEAD), history.json (full longitudinal replay), baseline.json (frozen anchor), REPORT.md, dashboard.html, and calibration/ (12-skill rubric + 3-skill functional eval) Key calibration finding: SQS measures structural/authoring quality, NOT usefulness (SQS<->FQD ~= -0.5). Three instruments triangulate quality: SQS (structure/regression), rubric review (content correctness), FQD (baseline-differential outcome). See quality/calibration/correlation.md.