feat(recce-dev): eval v2 framework + MCP impact_analysis improvements - #18

Merged

kentwelcome merged 12 commits into

mainfrom

feat/plugin-rename-and-recce-dev

Mar 31, 2026

Commits on Mar 24, 2026

feat(recce): stdio MCP transport + Option E hook cleanup

iamcxa
and
claude
committed
9134347
feat(recce-eval): add --skip-setup, --skip-teardown, --model flags

iamcxa
and
claude
committed
3fa7a56
fix(recce-eval): ch1-healthy-audit prompt — distinguish bugs from data patterns

iamcxa
and
claude
committed
cbb3508

Commits on Mar 25, 2026

feat(recce-eval): add reviewer workflow context for with-plugin runs

iamcxa
and
claude
committed
e3f1948
fix(recce-eval): skip MCP for fix-the-bug scenarios (skip_context)

iamcxa
and
claude
committed
614bd18
refactor(recce-eval): ch1-null-amounts from fix-the-bug to review-only

iamcxa
and
claude
committed
d90f9f0

Commits on Mar 30, 2026

feat(eval-v2): add foundation for dual-layer eval framework

authored and
kentwelcome
committed
8bf78e1
fix(eval-v2): redesign scenarios for existing-column diffs and dual-schema comparison

authored and
kentwelcome
committed
562bfb9
feat(eval-v2): add --mode flag to run-case.sh for fairness control

authored and
kentwelcome
committed
ee32912
docs(eval-v2): add v2 scoring rules to rubric

authored and
kentwelcome
committed
787f20a
fix(eval-v2): remove hardcoded workflow prompt from Mode C

authored and
kentwelcome
committed
c2759a7
feat(eval): add Spacedock scenario pipeline + R8 exclude-zero-orders scenario

authored and
kentwelcome
committed
3f8d551