Repository navigation
codedna v1.1.0
Prose-voice release. CodeDNA now profiles the words inside a repository as part of its fingerprint, while keeping repository evidence stronger than generic writing rules.
Added
- Measure comment length, sentence-like versus fragment-like style, capitalization, terminal punctuation, first-person and second-person use, and contractions.
- Measure capitalization and terminal punctuation in literal error messages for supported language patterns.
- Profile comments, errors, validation, logs, tests, CLI or UI text, and repository documentation as separate prose channels.
- Add positive
Voice cues to preserveguidance and a repository-aware prose fallback. - Add focused fixtures and regression coverage for every supported error-message pattern. The suite is 43 tests.
Changed
- Let Match and Check handle repository prose as well as code.
- Make generic prose checks subordinate to the profile and neighboring examples.
- Replace fixed report counts with the highest-signal findings supported by evidence.
- Update generated agent instructions, documentation, and installer metadata for v1.1.0.
Validation
- Python helpers compile successfully.
- All 43 unit tests pass.
- Skill validation passes.
- GitHub CI passes, including ShellCheck.
The prose-review concepts were adapted from pstack's unslop skill to CodeDNA's repository-evidence model.