Add GHCP end-to-end evaluation harness - #3072

Merged

George Ng (GeorgeNgMsft) merged 15 commits into

mainfrom

georgengmsft-ghcp-eval-implementation

Sep 30, 2026

Commits on Sep 24, 2026

Add credit-guarded GHCP evaluation harness and isolated workload

George Ng (GeorgeNgMsft)
and
Copilot
committed
cc3b1bb
Separate eval setup and grading and satisfy repository ratchets

George Ng (GeorgeNgMsft)
and
Copilot
committed
bb7c863
Correct frozen run metadata placement

George Ng (GeorgeNgMsft)
and
Copilot
committed
f1c332b
Explicitly pin mixed routing in isolated evaluation profiles

George Ng (GeorgeNgMsft)
and
Copilot
committed
73fc67f
Document measured GHCP evaluation blockers

George Ng (GeorgeNgMsft)
and
Copilot
committed
fc604fa
Repair scoped eval artifacts and scripted clarification continuation

George Ng (GeorgeNgMsft)
and
Copilot
committed
e847c7a
Close native no-replay gap for future GHCP eval runs

George Ng (GeorgeNgMsft)
and
Copilot
committed
4275a7c

Commits on Sep 25, 2026

Fix cross-platform GHCP evaluation artifact and fixture paths

George Ng (GeorgeNgMsft)
and
Copilot
committed
b47f8ba
Move Copilot evaluation harness into a dedicated package

George Ng (GeorgeNgMsft)
and
Copilot
committed
bb09c79
style: apply prettier formatting and policy fixes
typeagent-bot[bot]
committed
0b4c91b
Align evaluation README with normalized package metadata

George Ng (GeorgeNgMsft)
and
Copilot
committed
ba6a8b8
docs: regenerate README.AUTOGEN.md, command reference, and action browser
typeagent-bot[bot]
committed
0236149
Replace list-dependent eval cases with comparable file tasks

George Ng (GeorgeNgMsft)
and
Copilot
committed
45d45f0
Require file clarification before directory-backed native reads

George Ng (GeorgeNgMsft)
and
Copilot
committed
f2f045c

Commits on Sep 30, 2026

Merge main and preserve GHCP evaluation fallback controls

George Ng (GeorgeNgMsft)
and
Copilot
committed
460a64f