
Context Analyzer
- 29 repo stars
- Updated July 31, 2026
- jbdamask/john-claude-skills
Analyze Claude Code session JSONL to find what bloats the context window — image usage, file re-reads, tool-output volume — and get optimization recommendations.
About
Context-analyzer parses Claude Code chat-history JSONL files to identify what is consuming the context window — heavy image usage, repeated file reads, and large tool outputs. It then reports the worst offenders and recommends optimizations to keep sessions lean. Useful for power users diagnosing slow or expensive agent sessions.
- Parses JSONL session files
- Flags context-window bloat
- Reports image and re-read usage
- Optimization recommendations
Context Analyzer by the numbers
- Data as of Aug 1, 2026 (Skillselion catalog sync)
/plugin marketplace add jbdamask/john-claude-skills/plugin install context-analyzer@john-claude-skillsAdd your badge
Show developers this skill is listed on Skillselion. Paste this into your README.
| repo stars | ★ 29 |
|---|---|
| Last updated | July 31, 2026 |
| Repository | jbdamask/john-claude-skills ↗ |
What it does
Analyze Claude Code session JSONL to find what bloats the context window — image usage, file re-reads, tool-output volume — and get optimization recommendations.
README.md
context-analyzer
A Claude Code plugin that analyzes chat history to identify what's consuming your context window and provides optimization recommendations.
Features
- Context breakdown — Categorizes context consumption by type (images, file reads, bash output, screenshots, etc.)
- File re-read detection — Identifies files that are read repeatedly across sessions, wasting context
- Largest session analysis — Finds which sessions consumed the most context and why
- Actionable recommendations — Provides prioritized suggestions to reduce context bloat
- Auto-detection — Automatically finds the project history directory from your current working directory
Components
| Component | Name | Description |
|---|---|---|
| Command | /context-analyzer:analyze-context |
Run the full analysis and get a report |
| Agent | context-analyzer |
Subagent for autonomous context diagnostics |
| Skill | Context Optimization | Knowledge about context optimization patterns |
| Script | analyze_context.py |
Python analysis engine (no dependencies beyond stdlib) |
Installation
From a marketplace
/plugin install context-analyzer@your-marketplace
Local testing
claude --plugin-dir /path/to/context-analyzer
Usage
Slash command
/context-analyzer:analyze-context
Runs the analysis on the current project and presents a formatted report.
Agent (automatic)
The context-analyzer agent triggers when you ask questions like:
- "Why is my context filling up?"
- "Analyze my context usage"
- "What's consuming my context window?"
Direct script execution
# Markdown report (default)
python3 scripts/analyze_context.py ~/.claude/projects/-Your-Project-Path/
# JSON output for programmatic use
python3 scripts/analyze_context.py ~/.claude/projects/-Your-Project-Path/ --json
Prerequisites
- Python 3.8+ (uses only stdlib — no pip install needed)
- Claude Code chat history in
~/.claude/projects/
How It Works
The plugin parses JSONL chat history files that Claude Code stores in ~/.claude/projects/{encoded-project-path}/. Each line is a JSON message representing a conversation event. The analyzer categorizes every message into context consumption buckets and tracks file read patterns across all sessions.
See skills/context-optimization/references/analysis-methodology.md for details.