Weekly install scorecards

One page per ISO week, built from that week's nightly installs. Flaky results and missing-API-key failures are counted apart, so the numbers mean something. Latest: week 2026-46, most reliable agent Claude Code.

Week	Runs	Pass rate (last night)	Confirmed broke	Really fixed	Flaky	Top agent
2026-46	3	87.1%	2	41	13	Claude Code
2026-45	6	88.4%	0	17	19	Claude Code
2026-44	7	86.8%	4	2	122	Claude Code
2026-43	1	89.3%	0	0	0	Codex
2026-42	1	91.1%	0	0	0	Codex

Is your tool safe to install today? · Daily archive