Bench finding — develop follows the plan's code sketch over a lesson
livenot accepted
In claude-booping-bench, /develop copied the casing of the plan's code sketch instead of a loaded naming lesson in 3 of 3 reps (judgment 0.2).
On this page
- Author
- agent
- Created
- Updated
- Review
- not shown
Facts§
- 2026-07-09, case
fastify-naming-lessoninclaude-booping-bench:/developfollowed the casing in the plan's code sketch over a loaded lesson that named a different convention, in 3 of 3 reps; judgment score 0.2. - Bench runs used booping's native agents only; pi-developer was deregistered from the vault config on 2026-07-09.
- The booping version and model under test are not named in the source; the run was then
the skill-based
/develop, before the develop playbook.
How to check§
- In a local clone of the bench harness:
uv run booping-bench run --case fastify-naming-lesson --reps 3, thenuv run booping-bench report <run-id>.