Will I lose my code changes if I compact?

No, files saved to disk, Git state, and memory files persist. Compaction primarily clears the conversation history and intermediate reasoning to free up tokens.

Can I customize when the compaction suggestion appears?

Yes, you can adjust the COMPACT_THRESHOLD environment variable to set the specific number of tool calls that trigger a suggestion.

Why should I compact after a failed approach?

Compacting after a failure clears the 'dead-end' reasoning from the context, preventing the AI from being influenced by incorrect logic when attempting a new solution.

How does Strategic Compact differ from Claude's built-in auto-compaction?

Auto-compaction triggers at arbitrary token limits, which can interrupt mid-task logic. Strategic Compact suggests compaction at logical boundaries, such as after planning but before implementation.

Is this skill compatible with Cursor?

Yes, it is fully compatible. Note that in Cursor, the compaction command is /Summarize instead of /compact, but the strategic triggers remain the same.

Strategic Context Compaction

Name: Strategic Context Compaction
Author: flatrick

byflatrick

0•

Productividad y Flujo de Trabajo

Suggests manual context compaction at logical task boundaries to optimize performance and preserve critical reasoning during long AI coding sessions.

Strategic Context Compaction is designed to prevent the loss of critical information that often occurs during arbitrary automatic context compaction. By monitoring tool call volume and workflow transitions, this skill prompts users to run the /compact command (or /Summarize in Cursor) at specific milestones, such as moving from research to implementation. This ensures that bulky intermediate data is purged while essential distilled plans and persistent state are maintained, leading to faster, more coherent responses from Claude during complex, multi-phase development tasks.

Características Principales

010 GitHub stars

02Clear documentation on data persistence versus context loss

03PreToolUse hook integration for proactive workflow management

04Threshold-based suggestions triggered by tool call volume

05Logic-based compaction guidance for task phase transitions

06Customizable tool call thresholds via environment variables

Casos de Uso

01Clearing bulky debugging logs and research data before starting feature implementation

02Managing extremely long coding sessions that approach 200K+ token limits

03Maintaining AI focus when switching between unrelated tasks in a single session

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add flatrick/mdt strategic-compact

For use in Claude.ai and ChatGPT