Make hotword prompt wording more directive

The hotword injection copy now tells the model to prefer the user's spelling proactively while still avoiding unrelated mechanical replacements. That keeps the intent of issue #63 aligned with the existing prompt structure and only changes the wording that users see indirectly through polish behavior.

Constraint: Keep the change minimal and avoid changing the surrounding prompt architecture.
Rejected: Broaden the prompt or add new prompt sections | unnecessary for this issue.
Confidence: high
Scope-risk: narrow
Directive: Keep future hotword guidance user-facing and proactive, but still avoid blanket replacement language.
Tested: cargo test --manifest-path src-tauri/Cargo.toml polish::tests -- --nocapture; cargo check --manifest-path src-tauri/Cargo.toml
Not-tested: Live LLM output quality against the updated prompt wording.

openless-all/app/src-tauri/src/polish.rs

@@ -411,7 +411,7 @@ fn compose_system_prompt(mode: PolishMode, hotwords: &[String]) -> String {
         .collect::<Vec<_>>()
         .join("\n");
     format!(
-        "{}\n\n热词（用户提供的正确写法，仅当原始转写明显是其误识别时才纠正，不做机械替换）：\n{}",
+        "{}\n\n热词（用户希望以下写法在输出中保持准确；当转写中出现这些词的同音 / 近形误识别时，优先按上述写法输出，不做无关词的机械替换）：\n{}",
         base, bullets
     )
 }
@@ -888,4 +888,14 @@ mod tests {
         // 防回归：旧版"另外："标签写法不能再出现在示例输出里。
         assert!(!prompt.contains("另外：检查一下当前还有哪些 issues"));
     }
+
+    #[test]
+    fn compose_system_prompt_prefers_correct_spelling_for_hotwords() {
+        let prompt = compose_system_prompt(PolishMode::Light, &["GitHub".into(), "OpenLess".into()]);
+
+        assert!(prompt.contains("用户希望以下写法在输出中保持准确"));
+        assert!(prompt.contains("同音 / 近形误识别时，优先按上述写法输出"));
+        assert!(prompt.contains("- GitHub"));
+        assert!(prompt.contains("- OpenLess"));
+    }
 }