JMR-dev/Claude-Conversation-Exporter-TS
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

update readme to include note about long_conversation_reminder injections

Browse Source
This commit is contained in:
socketteer
2025-08-30 21:10:04 -07:00
parent 11796021b6
commit 930a3f8798
1 changed files with 3 additions and 3 deletions
Download Patch File Download Diff File Expand all files Collapse all files
README.md
+3 -3
View File
@@ -27,13 +27,13 @@ Claude.ai doesn't allow you to continue conversations after hitting context leng
- **Context compression** - Summarize earlier parts to fit more conversation
- **Selective context** - Choose which parts of the conversation to keep in context
### 3. **Escape Platform Restrictions**
Claude.ai uses a fixed system prompt that includes certain behavioral rules. Recent updates have added restrictions that some users find limiting, such as:
### 3. **Escape Platform Restrictions and "Long Conversation" Injections**
Claude.ai uses a fixed system prompt and injects "reminders" that include certain behavioral rules. Recent updates have added restrictions that some users find limiting, such as:
- Injunctions against Claude discussing its inner experiences or consciousness
- Specific formatting restrictions
- Behavioral constraints that may not align with all use cases
With exported conversations, you can continue them in environments with different or customizable system prompts.
With exported conversations, you can continue them in environments with different or customizable system prompts. Using the Anthropic API instead of Claude.ai also avoids "long_conversation_reminder" injections, though it doesn't avoid all injections.
### 4. **Enhanced Features in Other Apps**
Many third-party applications offer features not available on Claude.ai: