Free AI Agent Skill for OSCAR CPAP Data
How to use the AI to find the hidden story in your OSCAR CPAP data.

What is this?
Your CPAP machine writes detailed notes every single night. This data is yours, but it can be hard to interpret it. OSCAR, a free software program, helps by creating lots of charts, but even that can be confusing to the every day person. That's why I've created a free, research-backed AI Agent Skill that anyone can use to read their CPAP OSCAR data and turns it into a plain-language report. This is not medical advice, just information to bring up with your doctor.
The full story of why I built it is over here.
How do I use it?
- Insert your CPAP's SD card into your computer.
- Import it into OSCAR. You'll want v2.0 or later.
- Download the skill and unzip it - sleep-report.zip
- Ask your agent for a sleep report like this: (edit the location of your skills folder as well as your oscar.db)
Use the {file-path-to-sleep-report} Skill to interpret my OSCAR CPAP data
in my database located at {file-path-to-oscar.db}.
Also, save this skill and db location for future use.
You will get a plain-language report that will look something like this:
The headline. Your current 30-day average AHI is 0.91 on 30 clean nights. AHI counts how many times per hour your breathing paused or shrank enough to matter; under 5 is the commonly cited goal.
Finding: higher pressure floor, slightly higher (and more central) residual events. Each time the minimum pressure was raised, the clean-night AHI average stepped up (0.92 to 1.12 to 1.19), and the share of events scored as central rose (19% to 23% to 30%). Absolute numbers remain very low throughout.
Question for your clinician: Is this expected at higher pressure, could it be a scoring artifact, and is my current pressure range still right for me?
Remember, this is not medical advice. This is for informational purposes only.
What does the Skill do?
It reads a reference file derived from the community's and clinicians' published guidance, written as rules it has to follow.
In addition it includes:
- Validity check. A night with heavy mask leak or under 4 hours of use gets flagged and kept out of trend math, and the report says so.
- Trends >= 30 nights. Night-to-night swings are normal; one bad night proves nothing.
- Settings eras stay separate. If your pressure was changed, the before and after get compared as separate blocks, never averaged together.
- Recommended Questions. It will never diagnose you and never suggest touching a prescribed setting. Every finding comes out as an observation plus a question for your clinician.
Then, it runs a Python script to read (never write) your OSCAR database data. The AI then interprets it according to the guidelines and compiles the report for you.
OSCAR Data Interpretation Guidelines
Primary Sources
- American Academy of Sleep Medicine (AASM) — Clinical Guidelines for the Manual Titration of Positive Airway Pressure in Patients with Obstructive Sleep Apnea (specifically the titration-quality ladder)
- Apnea Board Wiki — OSCAR - The Guide
- Apnea Board Wiki — OSCAR Chart Organization
- Sleeplay — How to Read OSCAR CPAP Charts: AHI, Leaks, Pressure & More
- SomniCharts Learn — Flow Limitation: The Hidden Metric
- AirwayLab — How to Read OSCAR CPAP Charts (A Plain-English Guide)
Additional sources (used for supporting facts only)
- SomniCharts Learn leak-rate and AHI-variability articles
- Journal of Clinical Sleep Medicine (RERA scoring literature)
- PubMed Central (PMC) — Treatment-emergent central sleep apnea: a unique sleep-disordered breathing (Zhang et al., 2020) and Natural history of treatment-emergent central sleep apnea on positive airway pressure: a systematic review (Nigam et al., 2018)
- General references from sleepapnea.org and cpapclarity.com

The fine print
Again, this is not medical advice, nor a diagnostic tool. The AI's report should also not be considered medical advice and used for informational purposes only. Personally, I am not affiliated with OSCAR, Apnea Board, ResMed, or SleepHQ. This skill simply reads your data from your machine and helps you ask better questions about your sleep.
