Automated lab processing saves team 40 hours per month

A longevity clinic pulls dozens of new Quest Diagnostics lab results a day, and each one has to land in exactly the right patient's chart, correctly, once, with the right visibility. This Claude Skill runs that whole handoff process unattended through Cowork. It pulls Final results from Quest Diagnostics through the Quanum portal (physician.quanum.questdiagnostics.com), builds a male-specific bloodwork report taking the Quest HL7 file and turning it into something easy to read and understand for the patient through Modern Doc Media (biodesign.moderndocmedia.com), and matches each result to a patient chart in Simplified Practice (app.simplifiedpractice.com) using name and date of birth, not just a name. When a match isn't clean, it doesn't guess. There are safety checks for scenarios like a typo of a birth date or when someone uses a nickname for their lab. It goes further and checks the order # vs the lab # to confirm they match. In the event, there is still any question whether it is the right person, it uploads privately and flags it for a human in the loop so the patient can not see it until approved. Every step writes to a resumable ledger, so a login timeout or a browser crash costs nothing. The next run picks up exactly where the last one stopped, with zero risk of reprocessing a result that already made it through. In the event the site logs out and the user is unavailable for over 15min or longer, then the skill will upload the labs that it was able to download and covert, into the electronic medical records software to keep work flow moving ahead and not getting stuck waiting for the human in the loop. The output is a clean summary of successful male labs, female labs or "needs attention" posted straight to Slack that a staff member can act on in under a minute, not a wall of logs they have to audit by hand.

I have been through multiple iterations of the skill. I learned about drift and was able to reduce the skill to less than 500 lines to minis drift. The sites that Cowork interacts with tend to log out quickly, so I created a /skill-upgrader along with Fable to optimize the skill to 1. create a ledger in case the site does log out and we want to quickly and accurately be able to jump back in 2. figure out the right sequence of steps to efficiently get through the steps before sites begin logging out. It now processes everything much faster, with less tokens and close to no issues. Bottom line, this one automation is saving 2 members of my team a collective of about 40 hours per month.