compaction-audit

Audit compaction logs to summarize preprocessing steps and token changes.

Updated Apr 5, 2026
One-click install
npx skills add https://github.com/Lsogod/compaction-debugger --skill compaction-audit
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: compaction-audit
Source: https://github.com/Lsogod/compaction-debugger/tree/main/skills/compaction-audit
Command: npx skills add https://github.com/Lsogod/compaction-debugger --skill compaction-audit

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Audit a single compaction or preprocessing pass and answer why it was performed, what was compressed, and what was kept, enabling transparent explanations of compression decisions.

Core Features & Use Cases

  • Read and correlate logs from the compaction-debugger to determine whether preprocessing or compaction occurred.
  • Identify which messages were condensed into summaries and which recent messages were retained.
  • Verify whether a real /compact path was triggered and surface key metrics for debugging and parameter tuning.

Quick Start

Provide a session id or time range and I will audit the associated logs.

Frequently Asked Questions about compaction-audit

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I audit session logs to see what changed during a compaction step?

To audit compaction logs, provide a session id or time range. The skill correlates assemble, before-compaction, and after-compaction data to identify which messages were condensed and which recent messages were retained.

Why does my preprocessing pass retain recent messages while condensing older ones?

Preprocessing passes retain recent messages while condensing older ones to manage token usage. The skill audits these compaction decisions, summarizing why the step occurred and what specific items were compressed into summaries.

How can I verify if a real /compact path was triggered in my session?

You can verify a real /compact path by reading logs from the compaction-debugger. The skill compares before-compaction and after-compaction data to determine whether the path was taken and surfaces key metrics for debugging.

What do I need to provide to start a session analysis for compaction debugging?

You need to provide a session id or time range to start session analysis. The skill then reads the associated compaction-debugger logs to audit the preprocessing or compaction pass and return token changes with suggested parameter adjustments.

What key metrics are returned when diagnosing compaction issues?

When diagnosing compaction issues, the skill returns a concise conclusion with key figures including the session id, relevant file names, token changes, and suggested follow-ups or parameter adjustments for tuning.