strategic-compact

Suggest manual context compaction at logical phase boundaries in Claude Code sessions.

26|1|Updated Mar 19, 2026
One-click install
npx skills add https://github.com/BinyaminEden/claude-combine --skill strategic-compact-binyamineden
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: strategic-compact
Source: https://github.com/BinyaminEden/claude-combine/tree/main/skills/strategic-compact
Command: npx skills add https://github.com/BinyaminEden/claude-combine --skill strategic-compact-binyamineden

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Auto-compaction triggers at arbitrary points during Claude Code sessions, often mid-task, which discards critical context, breaks multi-step workflows, and ignores logical phase boundaries in your work, leading to lost progress and reduced output quality.

Core Features & Use Cases

  • Dual-signal suggestion system: Combines context size (scaled to your model's token window) and tool call count to trigger compaction prompts only when relevant, avoiding unnecessary interruptions.
  • Phase-aware guidance: Includes a decision table to help you identify optimal compaction points, such as after research, post-milestone, or before switching to unrelated tasks.
  • Configurable thresholds: Adjust tool call limits, context size triggers, and reminder intervals to match your workflow and model window size.
  • Use Case: For long development sessions, multi-phase projects (research → plan → implement → test), or when switching between unrelated tasks, this skill ensures you compact exactly when it won't disrupt your work, preserving relevant context for the current phase.

Quick Start

Activate the strategic-compact skill to receive timely suggestions for manual context compaction during long or multi-phase work sessions.

Frequently Asked Questions about strategic-compact

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I prevent auto-compaction from disrupting my Claude Code session workflow?

To prevent auto-compaction from disrupting your Claude Code session, trigger manual context compaction at logical phase boundaries using dual-signal suggestions. This approach monitors context size and tool call counts to prompt compaction safely between task phases.

What is context compaction and when should I trigger it during development?

Context compaction is the process of summarizing or clearing session history to manage token limits. You should trigger it during development at logical phase boundaries, such as after research, post-milestone, or before switching to unrelated tasks to preserve relevant workflow context.

How do I configure context size thresholds for token management in long coding sessions?

You configure context size thresholds by adjusting configurable parameters scaled to your model's token window. This allows you to set specific context size triggers and tool call limits that match your workflow and prevent unexpected interruptions during long coding sessions.

Does strategic compaction work with multi-phase development tasks like research and implementation?

Yes, strategic compaction works with multi-phase development tasks by providing phase-aware guidance. It includes a decision table to identify optimal compaction points, ensuring context is preserved when transitioning between research, planning, implementation, and testing phases.

Why does arbitrary auto-compaction cause context loss in multi-step workflows?

Arbitrary auto-compaction causes context loss because it triggers at fixed token limits regardless of task progress. This ignores logical phase boundaries, discards critical mid-task context, and breaks multi-step workflows, ultimately reducing output quality and losing progress.

Can I adjust tool call count thresholds to match my specific workflow optimization needs?

Yes, you can adjust tool call count thresholds alongside context size triggers and reminder intervals. These configurable thresholds allow you to match the compaction suggestion system to your specific workflow optimization needs and model window size.