How does the skill track when to suggest a compaction?

The skill uses a script that hooks into PreToolUse events, specifically counting 'Edit' and 'Write' tool calls. It suggests a compaction once it hits a set threshold.

Does this skill execute the /compact command automatically?

No, it only provides a suggestion. You decide whether or not to run the /compact command based on your current progress and task requirements.

Can I customize the frequency of suggestions?

Yes, you can set the COMPACT_THRESHOLD environment variable to change the number of tool calls that occur before the first suggestion is made.

Why use manual compaction instead of auto-compaction?

Auto-compaction triggers at arbitrary points and can purge important context mid-task. Strategic manual compaction allows you to choose logical boundaries, like after planning or debugging.

Strategic Context Compaction

Name: Strategic Context Compaction
Author: steven715

bysteven715

•

生産性とワークフロー

Optimizes Claude's memory management by suggesting manual context compaction at logical task boundaries to prevent data loss.

Strategic Context Compaction is a developer utility designed to give users more control over Claude's context window during long-running sessions. Rather than relying on arbitrary auto-compaction that might trigger mid-task and purge critical implementation details, this skill monitors tool usage and prompts the user to manually compact at strategic milestones. By suggesting compaction after exploration phases or before major architectural shifts, it ensures that Claude maintains a clean and relevant context, preserving essential project state while discarding transient research notes that are no longer needed.

主な機能

01Context preservation through logical task phases

021 GitHub stars

03Intelligent tool call tracking for Edit and Write operations

04Configurable suggestion thresholds via environment variables

05PreToolUse hook integration for seamless workflow prompts

06Periodic reminders to ensure context stays lean

ユースケース

01Clearing research and debugging context after identifying a fix but before writing code

02Managing large-scale refactors where preserving the implementation plan is critical

03Optimizing token usage during multi-step development cycles to prevent model degradation

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add steven715/cpp-claude-code strategic-compact

For use in Claude.ai and ChatGPT

Download Skill