What our sandbox result establishes
Our run on October 3, 2026, used commit c684cd4. The attached harness record reports a successful 15-second installation and a successful 6-second build step in our Debian sandbox. The recorded total is 29 seconds; adding the two step durations does not describe the whole run.
Tests were skipped because the harness detected no test script. A skipped test step is not a failed suite, and there are no passing-test counts to report. The build command returned zero, but its stored log tail is empty. That outcome does not establish that the Mac helpers compiled or that the device workflow ran.
The checkout contained 9 files and about 548 lines of source, occupying 0.2 MB. The record detected no CI workflows or tests directory. Pip-audit recorded 0 advisories, with packages recorded as 0 and no scannedPackages field. Scan coverage is unknown; the result does not establish that all 35 installed packages were scanned or that Airlift is secure.
The MrKeyoor testing methodology explains command outcomes and sandbox limits. That page provides provenance context, not evidence that its current procedure was used for this older run. This correction uses the stored record and includes no new execution.
What the current demonstration proves, according to its README
Airlift makes one new file outside AirTraffic's intended Media area. The current README describes a randomly named canary at a fixed /var/tmp destination on iOS 27.0 build 24A437, used on an iPhone 18 Pro. It moves the canary back into Media, reads it through AFC, and compares the recovered bytes.
The README explicitly limits the meaning of sandbox escape: the media service crosses its intended file boundary. It does not claim kernel access or control of the whole phone. Its destination list concerns fresh files, not overwriting existing ones, and a working path to the MobileGestalt plist has not been shown.
For interpreting a device result, exactBytesRecovered means the recovered bytes match the canary, while cleanupComplete means the temporary files were removed. The README says a successful run reports both as true. Returned AirTraffic API calls alone do not confirm delivery; file readback supplies evidence of the effect. If cleanup cannot finish, the documented result identifies a recoveryFiles directory on the Mac.
These are upstream descriptions of the current demonstration. Our sandbox did not verify them on hardware.
Keep the measured commit separate from current instructions
The README at our measured commit reported testing on iOS 27.0 RC build 24A435 and said final build 24A437 should also work. It described an interactive device selection flow and a configurable destination.
The current README instead documents cleanroom/poc.py, accepts build 24A437, and fixes the demonstration destination to /var/tmp. It attributes the other listed directory results to 24A435 and explicitly says this PoC does not exercise those directories on 24A437. Those current instructions and claims must not be presented as behavior tested by our October run.
Issue 4 reports a staging failure on 24A437 using the older commit c684cd41ca0ded2d1ab780c15f6ead05509ce062. Its author reports kAMDInvalidCheckinError and attributes the failure to an entitlement requirement. That is a reported failure in a specified environment. It neither proves that every 24A437 setup fails nor independently verifies the current cleanroom demonstration.
Before reproducing the result, record the checkout, Mac environment, iPhone model, and iOS build. Use that checkout's documentation, then inspect byte recovery and cleanup. Our successful sandbox command outcomes cannot substitute for those device checks.
Choose a device toolkit when the exploit is not your question
libimobiledevice is a cross-platform library for communicating with iOS device services without existing proprietary libraries. Its documented utilities cover tasks such as pairing, backups, diagnostics, and AFC access. Choose it for ordinary device-service integration; its documentation does not establish reproduction of Airlift's boundary escape.
pymobiledevice3 provides a Python API and command-line tool for interacting with iOS devices on Windows, Linux, and macOS. Choose it when Python automation and its documented device services answer your task. Either toolkit can complement a research workflow, but neither supplied source establishes it as a replacement implementation of this exploit.
How to read the retained ratings
These editorial scores use a five-point rubric: 1 indicates a research artifact with substantial validation gaps; 2 indicates substantial constraints or limited supporting evidence; 3 indicates adequate evidence for the stated use; 4 indicates clear, detailed support with remaining limits; 5 requires broad corroboration across the dimension. They are judgments, not measured performance or adoption.
Setup ease remains 2 because the meaningful workflow needs a Mac, Xcode, and paired hardware beyond our sandbox. Documentation remains 4 because the supplied README explains the boundary, prerequisites, result flags, and cleanup, while historical and current instructions require care. Community remains 2: the supplied issue provides a detailed external report, but the evidence does not establish response quality or adoption. Maturity remains 1 because our measured checkout had no detected tests or CI, and we have no independent device validation of the current demonstration.

