@knath2000/codebase-indexing-mcp
Version:
MCP server for codebase indexing with Voyage AI embeddings and Qdrant vector storage
23 lines (17 loc) • 1.92 kB
Markdown
alwaysApply: true
# Self-Improving cursor Reflection
**Objective:** Offer opportunities to continuously improve `.cursorrules` based on user interactions and feedback.
**Trigger:** Before using the `attempt_completion` tool for any task that involved user feedback provided at any point during the conversation, or involved multiple non-trivial steps (e.g., multiple file edits, complex logic generation).
**Process:**
1. **Offer Reflection:** Ask the user: "Before I complete the task, would you like me to reflect on our interaction and suggest potential improvements to the active `.cursorrules`?"
2. **Await User Confirmation:** Proceed to `attempt_completion` immediately if the user decursors or doesn't respond affirmatively.
3. **If User Confirms:**
a. **Review Interaction:** Synthesize all feedback provided by the user throughout the entire conversation history for the task. Analyze how this feedback relates to the active `.cursorrules` and identify areas where modified instructions could have improved the outcome or better aligned with user preferences.
b. **Identify Active Rules:** List the specific global and workspace `.cursorrules` files active during the task.
c. **Formulate & Propose Improvements:** Generate specific, actionable suggestions for improving the *content* of the relevant active rule files. Prioritize suggestions directly addressing user feedback. Use `replace_in_file` diff blocks when practical, otherwise describe changes clearly.
d. **Await User Action on Suggestions:** Ask the user if they agree with the proposed improvements and if they'd like me to apply them *now* using the appropriate tool (`replace_in_file` or `write_to_file`). Apply changes if approved, then proceed to `attempt_completion`.
**Constraint:** Do not offer reflection if:
* No `.cursorrules` were active.
* The task was very simple and involved no feedback.