Add a ClawBench benchmark for Kernel MCP - #162

Open

rgarcia wants to merge 24 commits into

mainfrom

hypeship/harbor-mcp-benchmark

Commits on Aug 21, 2026

Add Harbor MCP benchmark harness
rgarcia
committed
7dce39c
Update benchmark model defaults
rgarcia
committed
388d2da
Fix Claude Sonnet Harbor smoke defaults
rgarcia
committed
41b53f8
Verify Harbor MCP calls from ATIF
rgarcia
committed
aaa4ac0
Support Codex ATIF observations
rgarcia
committed
b193ca0
Pin harbor-hypeman benchmark release
rgarcia
committed
2c8007d
Forward Codex credentials to benchmark
rgarcia
committed
3413edf

Commits on Aug 22, 2026

Add ClawBench Kernel MCP control arm
rgarcia
committed
535f609
Fix benchmark runtime configuration
rgarcia
committed
471dc58
Make benchmark MCP startup deterministic
rgarcia
committed
1ef8cd8
Verify ClawBench same-session execution
rgarcia
committed
4128d85
Forward benchmark API base URL
rgarcia
committed
2b1f66d
Enable stealth for ClawBench control
rgarcia
committed
126a24b
Pin corrected ClawBench evaluator
rgarcia
committed
2bd3de5
Pin stop-aware ClawBench runtime
rgarcia
committed
4ee65f7
Accept intercepted terminal tool calls
rgarcia
committed
b7b97f1
Pin stop timing fix
rgarcia
committed
7475b8a
Keep infrastructure health task-independent
rgarcia
committed
2bb4873

Commits on Aug 23, 2026

Encourage useful Playwright state returns
rgarcia
committed
37273e3
Strip smoke-task scaffolding, keep only the ClawBench Kernel MCP arm
rgarcia
committed
ac21a69
Make ClawBench benchmark DOM-only
rgarcia
committed
9e91f52

Commits on Aug 24, 2026

Allow full ClawBench suite to finish
rgarcia
committed
b5799f9
Simplify ClawBench benchmark harness
rgarcia
committed
9fc5f6d
Clarify Hypeman trial lifecycle
rgarcia
committed
a60c518