strategic-compact

To preserve context in long-running AI assistant sessions, enable a skill that tracks message counts and suggests manual compaction at logical workflow breakpoints, keeping essential state and prompts intact.

You should compact context during multi-phase projects at logical workflow breakpoints, such as transitioning from research to plan, implement, or test phases, to maintain continuity and reduce context loss across tasks.

The stop-event hook tracks message counts during your session and uses a configurable threshold, defaulting to 50 messages, to emit non-blocking reminders to stderr that guide you to manually compact context.

Yes, the message threshold for compaction reminders is configurable. The skill uses a default threshold of 50 messages to trigger non-blocking suggestions at logical workflow breakpoints, which you can adjust for your environment.

Manual compaction does not disrupt frequent task switching because the skill emits non-blocking reminders to stderr. This guides you to compact context at logical breakpoints without interrupting your active workflow or session management.