Which tools trigger the compaction reminder?

The skill is specifically configured to monitor 'Edit' and 'Write' tool calls, as these typically represent significant progress in a development task.

Does this skill run the /compact command automatically?

No, it only suggests that you run it. This keeps the user in control, allowing you to decide if the current context is still necessary for your immediate task.

Why is strategic compaction better than auto-compaction?

Auto-compaction can trigger at any moment, often mid-task, which might cause Claude to lose focus on your current goal. Strategic compaction lets you choose logical stopping points to refresh the context window.

How do I adjust the suggestion frequency?

You can set the COMPACT_THRESHOLD environment variable. By default, it suggests compaction after 50 tool calls and reminds you every 25 calls after that.

Strategic Context Compaction

Name: Strategic Context Compaction
Author: arabicapp

byarabicapp

•

生产力与工作流

Optimizes Claude's token usage by suggesting manual context compaction at logical task boundaries rather than arbitrary intervals.

Strategic Context Compaction is a workflow optimization utility designed to prevent the loss of critical context during complex coding sessions. While Claude's native auto-compaction triggers at arbitrary token limits, this skill monitors tool usage and prompts you to manually compact history at logical milestones—such as after finishing a research phase or completing a major debugging cycle. By providing these strategic reminders through PreToolUse hooks, it ensures that your most relevant implementation plans and logic are preserved exactly when you need them most, leading to higher accuracy in long-running agentic tasks.

主要功能

01Environment variable configuration for flexible usage

02Intelligent tool-call tracking for Edit and Write operations

03Customizable thresholds for compaction suggestions

04PreToolUse hook integration for seamless workflow monitoring

05Strategic reminders to preserve context during multi-step tasks

061 GitHub stars

使用场景

01Preserving an implementation plan after a long research and exploration phase

02Cleaning up error-resolution context after a successful debugging session

03Managing long-running sessions to prevent arbitrary mid-task context loss

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add arabicapp/everything-claude-code strategic-compact

For use in Claude.ai and ChatGPT

Download Skill