Check for duplicate issues before filing

Use this before filing a new bug issue, to surface candidate duplicates via keyword search.

Steps

1. Call with a short, keyword-heavy query describing the defect — not the full draft title or body:

   search_similar_issues {
     owner: "<owner>",
     repo: "<repo>",
     query: "crash save filename slash"
   }

2. Read the response:

   {
     "candidates": [
       { "number": 88, "title": "App crashes saving files with slashes", "state": "open", "htmlUrl": "https://github.com/..." }
     ],
     "totalCount": 1
   }

3. Review each candidate’s title and state yourself. This is a plain keyword search against GitHub’s search/issues REST endpoint — it does not rank by semantic similarity, so a candidate appearing here is not a confirmed duplicate, and a real duplicate phrased very differently may not appear at all.

4. If you find a genuine duplicate among the candidates, file nothing new; if the new report is itself the duplicate, see close-as-duplicate.md.

Notes

• This tool only searches; it never files or closes anything.
• AI/embedding-based similarity is explicitly out of scope for this tool per the plugin’s originating research report — don’t expect matches based on meaning rather than shared keywords.

See also

• reference/tools.md — full input/output schema.
• how-to/close-as-duplicate.md — once you’ve confirmed a duplicate.