What happens to my work when I compact?

Compaction clears conversation history and intermediate reasoning, but files on disk, CLAUDE.md instructions, TodoWrite task lists, and Git state remain intact.

When is the best time to compact?

The best times are after finishing a research phase, once a plan is written to a file, or after a successful debugging session before starting new work.

How do I configure the tool call threshold?

You can set the COMPACT_THRESHOLD environment variable in your terminal to change the number of tool calls that occur before the first suggestion (default is 50).

Why use manual compaction over auto-compaction?

Auto-compaction can trigger at arbitrary points, often mid-task, causing Claude to lose track of variable names or logic. Strategic compaction ensures you only clear history at logical milestones.

Does this skill work automatically?

It provides suggestions based on your activity via hooks, but the decision to execute the /compact command remains with the user to ensure context is never lost unexpectedly.

Strategic Context Compaction

Name: Strategic Context Compaction
Author: saar120

bysaar120

0•

生产力与工作流

Suggests manual context compaction at logical task boundaries to optimize token usage and prevent loss of critical project state.

Strategic Compact is a specialized Claude Code skill designed to help developers manage long-running sessions effectively. Instead of allowing Claude to perform arbitrary auto-compaction—which often occurs mid-task and leads to the loss of critical reasoning or state—this skill monitors tool usage and prompts the user to perform a manual /compact command at strategic milestones. By aligning context clearing with logical transitions, such as moving from research to implementation or clearing debugging logs before a new feature, users can maintain high response coherence and speed without sacrificing the context needed for complex operations.

主要功能

01Preservation of critical state through milestone-aware clearing

02Configurable reminder frequency via environment variables

03PreToolUse hook integration for non-intrusive workflow alerts

040 GitHub stars

05Threshold-based compaction reminders after 50 tool calls

06Logical phase-transition guidance for multi-step tasks

使用场景

01Managing performance in long-form sessions exceeding 200K tokens

02Clearing bulky error logs and debug traces after resolving a complex bug

03Transitioning from broad research and planning to focused code implementation

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add saar120/skillgate-registry strategic-compact

For use in Claude.ai and ChatGPT

主要功能

01Preservation of critical state through milestone-aware clearing

02Configurable reminder frequency via environment variables

03PreToolUse hook integration for non-intrusive workflow alerts

040 GitHub stars

05Threshold-based compaction reminders after 50 tool calls

06Logical phase-transition guidance for multi-step tasks

使用场景

01Managing performance in long-form sessions exceeding 200K tokens

02Clearing bulky error logs and debug traces after resolving a complex bug

03Transitioning from broad research and planning to focused code implementation

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add saar120/skillgate-registry strategic-compact

For use in Claude.ai and ChatGPT