What problem does it solve?
Context-engine prevents long AI coding sessions from losing critical decisions, progress, and behavioral mode context when the window fills up or compaction becomes necessary.
Core Features & Use Cases
- Context health monitoring: Tracks tool-call pressure and classifies sessions as green, yellow, orange, or red.
- Safe compaction support: Preserves task state, file history, decisions, and continuation points before context is compacted.
- Behavioral mode injection: Loads dev, research, or review guidance from context files so orchestrators can shift session behavior without rewriting rules.
- Output control and preview gating: Activates terse output when needed and estimates bundle cost before expensive multi-file dispatches.
- Use Case: A long implementation session reaches its limit, so the skill saves state, recommends compaction at the next boundary, and keeps the work resumable.
Quick Start
Use the context-engine skill to assess current session pressure, preserve state if compaction is needed, and recommend the safest next boundary for continuing work.