reuse-scanlisted
Install: claude install-skill taipt1504/claudehut
# Reuse-Scan
ClaudeHut không tự build phân tích — detect + invoke native + normalize.
## Quick start
1. Run `scripts/detect-integrations.sh` — ghi `state/integrations.json`.
2. Based on integrations:
- UA available → invoke `/understand-chat "<topic + nouns>"`.
- Graphify available → `Bash: graphify query "<topic>"`.
- Both → invoke parallel, merge by path.
- Neither → `scripts/reuse-scan-grep.sh <topic>` (fallback).
3. `scripts/normalize-candidates.sh` → top-5 normalized list.
4. Write `state/tasks/<task-id>/reuse-scan.json` với timestamp.
5. Present candidates kèm `reuse | adapt | refuse` prompt.
## Backend matrix
Detailed: `references/backend-matrix.md`.
| Integration available | Native command invoked | Source weight |
|------------------------|------------------------|---------------|
| understand_anything | `/understand-chat "<topic>"` HOẶC parse `knowledge-graph.json` | 1.0 |
| graphify | `Bash: graphify query "<topic>"` | 0.9 |
| graphify global | `Bash: graphify global query "<topic>"` | 0.9 + 0.2 cross-project |
| neither | grep + heuristic | 0.7 |
## Output schema
Each candidate normalized:
```json
{
"path": "src/main/java/com/x/UserService.java",
"class": "UserService",
"purpose_one_line": "CRUD + duplicate-check",
"score": 0.92,
"source": "ua|graphify|graphify_global|grep",
"layer": "Service|Repository|Controller|Util",
"cross_project": false
}
```
Ranking heuristics: `references/ranking-heuristics.md`. Schema details: `referen