Deploying agents inside a firm is an operations problem before it is a model problem. These are the tools we built to keep our own work reviewable, measurable, and consistent — released so your engineers can hold their agents to the same bar.
A firm that runs agents needs the same things it needs from an analyst: reviewable work, measurable quality, and consistent method across everyone doing the job.
Everything here is in use in the AI deployments we run. We built these tools because we needed them, they hold up under real client work, and they are the parts every serious deployment needs — so we publish them.
A useful skill gets copied. One plugin becomes twenty — and every copy needs maintaining.
Six months ago you copied a skill into five plugins. Today you fix a bug in one of them. The PR looks correct, CI is green, it ships — and the bug is still live in the other four, because nothing in the review told you they were related. 100xdrift-check reads the change, finds those copies, and comments on which ones need the same fix.
Copies are allowed to diverge. Drift Check looks for shared ancestry and purpose, not identical text.
It never rewrites files, propagates changes, or blocks a merge. It reports; you decide.
“Does this fix need to propagate anywhere else before we merge?” Asked at review time, not after.
It is early. We especially want to hear where its sibling detection gets things wrong — open an issue.
Open tools give your engineers a shared standard. Turning your firm's data and judgment into agents that hold up in production is what we do with you.