Inspect MuseScore archives with AI
Follow a MuseScore 4 file through scorelib, distinguish preserved archive entries from regenerated notation and check an observed conversion loss.
Inspect MuseScore archives with AI when a successful file open hides a musical change. A repeat sign may survive in the original XML but vanish from a score rebuilt from an intermediate representation. The gp-mscz-format skill helps contributors to scorelib follow those two routes independently.
An MSCZ archive can contain more than notation. A thumbnail or a style file might pass straight through without becoming part of the simplified music model. If the task is to preserve a musical detail, compare that detail at each stage and consult the loss report where the conversion cannot represent it.
What it does
- Identify the main score XML inside an archive.
- Separate raw-file preservation from a regenerated score.
- Compare repeat counts, ties or string numbers across a conversion.
- Check the unit used for tempo in each representation.
- Record an unsupported detail next to the test that exposed it.
Install and use
- Provide the agent with scorelib, a Rust toolchain and an authorised MuseScore 4 fixture. Identify a measure or track with a known feature.
- Ask, “The repeat at bar eight disappears. Does it survive parsing, and does the rebuilt score include it?”
- Run the relevant local tests and compare the specific measure before treating a successful open as a complete result.
FAQ
Source
Made by slundi. Original skill: gp-mscz-format, published on Codeberg under the MIT licence.
View on Codeberg