compaction-tuner

Tune compaction parameters by analyzing conversation logs and user goals to adjust values like keepRecentMessages and summaryCharBudget, balancing token savings with essential context retention. The skill outputs recommended configurations with rationales.

Preventing over-compression in long logs involves adjusting maxSummaryBullets and summaryCharBudget based on log analysis. The skill provides tuning strategies ranging from conservative to aggressive to maintain stability across varying conversation lengths.

Configure keepRecentMessages and maxSummaryEntries by starting with a conservative tuning setup for daily conversations. The skill provides a default quick-start configuration and explains the rationale behind each parameter's value to ensure stability.

Yes, you can save tokens by enabling dropOldToolMessages and dropTrivialAssistantMessages. The skill evaluates your logs to recommend whether dropping these specific message types suits your context retention goals.

Adjustable parameters for compaction-debugger log analysis include keepRecentMessages, maxSummaryEntries, summaryCharBudget, maxSummaryBullets, dropOldToolMessages, and dropTrivialAssistantMessages.

Use aggressive compaction tuning strategies when maximizing token savings is the priority and minor context loss is acceptable, whereas conservative strategies suit scenarios requiring strict long-term stability and context preservation across varying conversation lengths.