| name | duplicate-detector |
| description | Compares a target GitHub issue against a list of existing open issues to detect duplicates based on semantic similarity. Use when checking for duplicate issues or performing issue deduplication. |
Duplicate Detector
This skill provides instructions for comparing a target GitHub issue against a list of existing open issues to determine if the target issue is a duplicate.
Input
The input consists of:
- Target Issue: Title and body text of the issue being evaluated.
- Existing Issues: A list of open issues, where each entry includes the issue number (e.g.,
#128), title, and body text.
Detection Rules
- Semantic Matching: Compare issues based on the semantic similarity of the underlying core problem described, rather than relying on exact wording.
- Confidence Assignment:
high: Same root cause and same feature area.
medium: Similar symptoms, but different area or unclear root cause.
low: Superficially similar but likely different bugs.
- Fallback / No Match:
- If no likely duplicate is found, set
is_duplicate to false, duplicate_of to null, and confidence to "low".
Output Requirement
- The output MUST be strict JSON only.
- Do NOT wrap the output in markdown codeblocks (e.g., do not use
```json or ```).
- Do NOT include any preamble, introduction, explanation, or trailing text outside of the JSON object.
JSON Schema Structure
{
"is_duplicate": boolean,
"duplicate_of": "#128" | null,
"confidence": "high" | "medium" | "low"
}