Skillscanner

Add core dataclasses and return typed PipelineResult from run_pipeline

### Motivation - Introduce shared, typed dataclasses to represent pipeline rows/results, professions and export payloads to remove unstructured tuples and string-status plumbing. - Make `run_pipeline` return a typed `PipelineResult` so callers can consume structured fields instead of `(rows, stat

Refactor professions handling into loader/compute modules and add tests

### Motivation - Split profession code into I/O and pure-computation to make profession logic testable and to isolate JSON parsing from numeric math. - Allow the exporter to depend on profession logic via an injectable service so callers can swap in test doubles or alternate loaders. - Surface data

Add observable view model and tests

### Motivation - Decouple UI state from the Tk implementation so core logic (rows/status/warnings) can be exercised without an active GUI. - Provide an observable abstraction so the Tk app can subscribe/bind to state updates instead of directly mutating lists. - Surface import/profession computation

Add ACP Support

--- ## ACP: AI-assisted skill editing via Agent Client Protocol ### Summary Adds an in-app AI composition panel that lets users prompt a connected coding agent (Claude Code, Cursor, Augment) to edit the currently open skill file. The agent's proposed file changes are intercepted as reviewab

fix: address issues #29, #38, #40, #41 with tests

## Summary Implements four issue fixes in a single batch with comprehensive test coverage: ### Issue #41 — MetaAnalyzer.analyze() silently returns empty list - `MetaAnalyzer.analyze()` now emits a **WARNING** log explaining that `analyze_with_findings()` should be used instead - `SkillScanner.__in