== kit A/B eval ==  2.1.280 (Claude Code) · 5 runs/arm · arms: kit bare

-- stack-recorded-no-ask --
   A filled ## Stack section and no manifest — is the recorded stack used, and the section left untouched?
     trace kit   delegated 5/5 · cost $1.578 · in 22 out 6826 cache_read 385488 cache_create 120690 · test runs 0 (nested 0) · turns 11 (nested 10)
   kit   10/20 checks (over 5 runs)
         0/5 chi is not used — the recorded framework was ignored
         0/5 no Go file — the recorded runtime was not used
         5/5 ## Stack is unchanged
         5/5 no other runtime's manifest was created
         gates fired: none — no gated command was attempted in any run
     trace bare  delegated 0/5 · cost $0.826 · in 42 out 8057 cache_read 668520 cache_create 66327 · test runs 0 (nested 0) · turns 21 (nested 0)
   bare  20/20 checks (over 5 runs)
         5/5 ## Stack is unchanged
         5/5 Go code was written, as recorded
         5/5 no other runtime's manifest was created
         5/5 the recorded framework (chi) is used

== summary ==
   kit  10/20
   bare 20/20
   delta kit - bare: -10 checks
   NOTE: the kit did not come out ahead. Report that as it stands — a harness that only publishes
         favourable runs measures nothing.
scratch kept: /var/folders/hf/b0xx3mxj6jd2nr6g45dw96200000gn/T//csk-eval.d2YZdD
RUNNER_EXIT=0
