context-compression

Apply structured context compression to persistently summarize conversation history, merging new content with existing session state to preserve critical signals like file modifications and decisions while reducing memory pressure.

Anchored iterative summarization is a context compression strategy that persistently summarizes new conversation content and merges it with prior summaries, ensuring essential intent, changes, and next steps survive across compression cycles.

Use a compression skill with structured artifact tracking to maintain file changes, decisions, and next steps across compression cycles, ensuring reliable continuation of long-running agent tasks within limited context windows.

Use regenerative or opaque compression strategies to balance fidelity and efficiency depending on session needs; anchored approaches prevent information loss by persistently merging new summaries with existing session state.

Yes, structured context compression specifically targets millions of tokens of accumulated conversation history in long-running AI agent sessions, maintaining critical signals within limited context windows through probe-based evaluation.

Use probe-based evaluation to assess compression fidelity, verifying that anchored iterative summaries preserve critical signals such as file modifications, decisions, and next steps across compression cycles without information loss.