| name | unused-avalonia-styles |
| description | Scan Avalonia `.axaml` style selectors for likely unused or mismatched styles by comparing `Style Selector="..."` patterns against XAML and C# usage evidence. Use when the user asks to find dead Avalonia styles, stale selector branches after control/icon migrations, or likely-unused style classes/names instead of reviewing selectors manually. |
Unused Avalonia Styles
Use this skill to produce a high-signal report of likely dead Avalonia selectors. The scanner is heuristic: it finds strong cleanup candidates and ambiguous selectors that need manual review, but it does not prove safety for removal.
Execution
- Run:
python .codex/skills/unused-avalonia-styles/scripts/run_unused_avalonia_styles.py --workspace-root .
- Read
.artifacts/unused-avalonia-styles/unused-summary.md first.
- Use
.artifacts/unused-avalonia-styles/unused-summary.json only when structured post-processing is useful.
What The Scanner Checks
- Reads
.axaml files and extracts Style Selector="..." values.
- Collects XAML usage evidence from element tags plus
Classes, Name, and x:Name.
- Collects simple code-behind evidence from
new ControlType, .Classes.Add("..."), and .Name = "...".
- Compares anchored selector components such as
Path.some-class, #SomeName, or Button.foo against that evidence.
Output Buckets
actionable: anchored selector components have no matching evidence; these are strong cleanup candidates.review: selectors use syntax the scanner does not model well, or are broad plain-type selectors where generated/template usage may exist.
Limits
- Treat the report as
likely unused, not as proof.
- Descendant selectors are checked component-by-component; the scanner does not build a full visual-tree matcher.
- Template-generated controls can make broad type-only selectors look unused, so those are routed to
review.
- Reflection, runtime class toggles, or highly dynamic control construction can evade the scan.
Response Shape
When reporting results:
- State the exact workspace scanned.
- Link the markdown and JSON report paths.
- Present
actionable candidates first, then review items.
- Call out the heuristic nature of the scan when recommending deletions.
