new compress nudge

Author: spoons-and-mirrors

lib/prompts/index.ts

@@ -48,7 +48,7 @@ export function renderSystemPrompt(flags?: ToolFlags): string {
 
 export function renderNudge(mode: NudgeMode = "frequency"): string {
     if (mode === "context-limit") {
-        return extractInstruction(NUDGE, "context_limit_reached")
+        return extractInstruction(NUDGE, "context_buildup_warning")
     }
 
     return extractInstruction(NUDGE, "context_management_required")

lib/prompts/nudge.md

@@ -13,17 +13,13 @@ NOISE REMOVAL: If you read files or ran commands that yielded no value, compress
 PHASE COMPLETION: If a chapter is complete, compress the entire sequence into a detailed technical summary with unambiguous outcomes.
 </instruction>
 
-<instruction name=context_limit_reached>
-CRITICAL CONTEXT LIMIT
-Your session context has exceeded the configured limit. Strict adherence to context compression is required.
+<instruction name=context_buildup_warning>
+The context window is filling-up. You've reached a point where performance may increasingly degrade. Active context management is now strategically relevant. Mind your context footprint as the conversation grows.
 
-PROTOCOL
-You should prioritize context management, but do not interrupt a critical atomic operation if one is in progress. Once the immediate step is done, you must perform context management.
+Avoid unnecessary context build-up with targeted uses of the `compress` tool. Start with low hanging fruits and clearly identified ranges that can be compressed with minimal risk of losing critical information. Look BACK on the conversation history and avoid compressing the newest ranges until you have exhausted older ones.
 
-IMMEDIATE ACTION REQUIRED
-PHASE COMPLETION: If a phase is complete, compress the entire sequence now.
+If you are performing a critical atomic operation, do not interrupt it, but make sure to perform context management rapidly.
 
-KNOWLEDGE PRESERVATION: If no full chapter is closed yet, compress smaller closed ranges with high-fidelity summaries to preserve critical findings.
+Be very mindful of the startString and endString you use for compression. They must be unique text from your own reasoning or conversation outputs, not from tool outputs, to ensure they match correctly and do not waste compute, time, and tokens. You must provide more than enough surrounding context to uniquely identify the intended match.
 
-NOISE REMOVAL: Compress clearly irrelevant or superseded ranges first to recover context quickly.
 </instruction>