yuzhe Claude commited on
Commit
d5f5aec
·
1 Parent(s): 944a301

Add usage principles for developer prompt

Browse files

- Add usage principles section with clear rules
- Emphasize placing developer prompt only once at session start
- Clarify not to include in user/assistant messages
- Provide correct usage pattern example

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>

Files changed (1) hide show
  1. README.md +23 -0
README.md CHANGED
@@ -235,6 +235,29 @@ User: "Search for SOL on Solana" Model:
235
 
236
  For optimal performance, use the following developer/system prompt when initializing the model:
237
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
238
  ```json
239
  {
240
  "messages": [
 
235
 
236
  For optimal performance, use the following developer/system prompt when initializing the model:
237
 
238
+ ### Usage Principles (Important)
239
+
240
+ **Follow these rules for best results:**
241
+
242
+ 1. **Place Once at the Beginning**: Put the developer prompt only once, at the very start of your conversation session
243
+ 2. **Do NOT place in user messages**: Never include the developer prompt content in user/assistant messages or tool schemas
244
+ 3. **Session-wide persistence**: For multi-turn conversations, keep the same developer prompt at the session start - do not repeat it
245
+
246
+ **Correct usage pattern:**
247
+ ```json
248
+ {
249
+ "messages": [
250
+ {"role": "developer", "content": "<developer prompt goes here>"},
251
+ {"role": "user", "content": "first user query"},
252
+ {"role": "assistant", "content": "assistant response"},
253
+ {"role": "user", "content": "second user query"}
254
+ // No need to repeat developer prompt in subsequent turns
255
+ ]
256
+ }
257
+ ```
258
+
259
+ ### Developer Prompt Content
260
+
261
  ```json
262
  {
263
  "messages": [