steventsvik/tellurion is a monitoring app for watching in real time how AI coding agents such as Claude Code, Codex, and Gemini work in a single macOS terminal. PR #118, merged on November 4, 2026, fixed the release check side. It eliminated the approach in which release.sh hardcoded the names of five test-only switches to look for in the built app binary, and replaced it with a new script, release-checks.sh, that extracts the current names directly from the Swift source and compares them. In the same commit, release.sh also changed how it checks the upstream version, reading the package version in build.zig.zon instead of using git tags, and one line was added to the CI workflow to run the newly added test_release_checks.py. The change is 218 lines: 216 additions and 2 deletions, affecting four files.
Here, a "switch" means an internal environment variable or flag that goes only into test builds. This app has to filter out binaries that still contain test hooks before shipping, but if the list of names is written out by hand, once the names change in the source the grep finds nothing and the check can appear to have passed. The new script reads names in the form TestHooks.value or TestHooks.flag from the source and takes names beginning with a capital letter as candidates, and because of the limitation that Swift hides strings of 15 bytes or fewer inside commands so that grep cannot see them, it treats only names of 16 bytes or more as check targets. There are three return values. 0 means clean, 1 means a switch was found, and 2 means the check itself could not be run, with 2 also treated as a failure. The principle of not treating a situation in which the check could not be run as a pass is written into the code comment as-is.
That said, this check is a string search that depends on naming rules and the placement of the source. If fewer than 20 names are extracted from the source, that alone is reported as a failure, and if the length and spelling of the names change, the search target changes accordingly. The new test directly creates a file for passing and a file for failing and runs the script to verify it, but there are no execution result figures in the commit materials. PR #118 is merely a change that refines an internal safety mechanism in the release pipeline, and there is nothing in the materials showing changes to application features or performance. It appears to be a change aimed at improving the accuracy of the release flow, but it is still too early to judge the actual release results.