Last night’s test run
Every skill ships with test cases. They run every night against the current model and the result is written down, pass or fail. This is that result, straight from the run. Last run finished Aug 24, 2026, 2:56 AM CT.
49 of 50 passedacross 5 skills
Model: claude-opus-5[1m]. Cases per skill are split between prompts that should trigger the skill and prompts that should not, so a skill that fires on everything fails too.
Skill
Result
What missed
client-brief (starter pack)
10 of 10
Nothing missed.
follow-up-draft (starter pack)
10 of 10
Nothing missed.
meeting-prep (starter pack)
10 of 10
Nothing missed.
proposal-outline (starter pack)
10 of 10
Nothing missed.
weekly-review (starter pack)
9 of 10
wr-06: missing: ['risk']
1 cases missed last night
A miss opens a fix task the same night. The fix ships as a pack update, and the update note says what changed and why. Nothing is quietly re-run until it passes. If a skill stays down, the badge on the home page stops showing numbers rather than showing you a good night.