When is the compaction process triggered?

Compaction typically triggers when token usage hits 75-90% of the limit, with warning preparations starting at 50% and aggressive compaction occurring above 90%.

Will I lose critical project details during compaction?

No, the tool follows a strict 'Must Preserve' strategy that ensures architectural foundations, error patterns, and current feature states are never deleted.

Does this skill help reduce AI costs?

Yes, by reducing the total token count in the context window, it lowers the cost per request and prevents unnecessary token expenditure on stale logs.

How does Context Compactor decide what information to keep?

It uses an importance scorer that prioritizes architectural decisions, unresolved bugs, and recent file changes while summarizing completed tasks and discarding redundant tool outputs.

Context Compactor

Name: Context Compactor
Author: adaptationio

byadaptationio

Productivity & Workflow

Optimizes AI token usage by automatically summarizing long-running sessions while preserving critical architectural and state information.

About

Context Compactor is a specialized utility designed for Claude Code to manage high-volume sessions that approach token limits. It employs an intelligent preservation strategy that distinguishes between 'must-keep' data—such as architectural decisions and unresolved bugs—and 'discardable' data—such as verbose logs and stale search results. By scoring message importance and validating coherence post-compaction, it ensures that Claude remains efficient and context-aware without hitting memory boundaries, making it essential for complex, multi-stage development tasks where session longevity is key.

Key Features

0 GitHub stars
Automated summarization of completed features and resolved errors
Preservation of critical architectural decisions and error patterns
Token budget monitoring with multi-tier action triggers
Coherence validation to ensure compacted context remains logical
Intelligent importance scoring for message prioritization

Use Cases

Streamlining context for project handoffs between different AI agents or sessions
Managing long-running autonomous coding sessions that span hundreds of messages
Optimizing token budgets to reduce latency and operational API costs

About

Key Features

0 GitHub stars
Automated summarization of completed features and resolved errors
Preservation of critical architectural decisions and error patterns
Token budget monitoring with multi-tier action triggers
Coherence validation to ensure compacted context remains logical
Intelligent importance scoring for message prioritization

Use Cases

Streamlining context for project handoffs between different AI agents or sessions
Managing long-running autonomous coding sessions that span hundreds of messages
Optimizing token budgets to reduce latency and operational API costs