compact
About
This skill manually triggers context saving in VSCode/Cursor extensions when auto-compact fails, allowing you to preserve state before starting fresh. Use it when context reaches 80%+ usage and you need to maintain conversation history. It provides a workaround for limited extension hook support by giving developers direct control over context management.
Quick Install
Claude Code
Recommended/plugin add https://github.com/majiayu000/claude-skill-registrygit clone https://github.com/majiayu000/claude-skill-registry.git ~/.claude/skills/compactCopy and paste this command in Claude Code to install this skill
Documentation
Manual Context Save (Compact Workaround)
Why This Exists
VSCode and Cursor extensions have limited hook support (GitHub issue #15021).
The PreCompact hook may not trigger automatically, leading to lost context.
This skill provides manual control to save state before context overflow.
When to Use
- Context warning shows 80%+ usage
- Working in VSCode or Cursor extension
- Before starting a new conversation
- When auto-compact doesn't seem to work
Execution Steps
Step 1: Detect Environment
# Check current environment
source ~/.claude/hooks/detect-environment.sh 2>/dev/null && print_env_info || echo "Environment detection unavailable"
Step 2: Get Current Session ID
The session ID is needed to save context properly:
# Try to get session ID from state
SESSION_ID=$(cat ~/.ralph/state/current-session 2>/dev/null || echo "manual-$(date +%Y%m%d-%H%M%S)")
echo "Session: $SESSION_ID"
Step 3: Extract and Save Context
# Run the pre-compact hook manually
export SESSION_ID="${SESSION_ID:-manual-$(date +%Y%m%d-%H%M%S)}"
# Create input JSON for the hook
echo "{\"hook_event_name\":\"PreCompact\",\"session_id\":\"$SESSION_ID\",\"transcript_path\":\"\"}" | \
~/.claude/hooks/pre-compact-handoff.sh
echo ""
echo "✅ Context saved to:"
echo " Ledger: ~/.ralph/ledgers/CONTINUITY_RALPH-$SESSION_ID.md"
echo " Handoff: ~/.ralph/handoffs/$SESSION_ID/"
Step 4: Verify Save
# Show the saved ledger
echo "=== SAVED LEDGER ==="
head -30 ~/.ralph/ledgers/CONTINUITY_RALPH-$SESSION_ID.md 2>/dev/null || echo "Ledger not found"
Alternative: Use Ralph CLI
If the above doesn't work, use the Ralph CLI directly:
ralph compact
This wrapper command handles all the complexity automatically.
Post-Compact Actions
After saving context:
- Start fresh: Use
/clearor start a new conversation - Restore context: The
SessionStarthook will auto-load the saved ledger - Verify restoration: Check that your objective is loaded
Recovery
If context was lost without saving:
# List recent ledgers
ralph ledger list
# Load a specific ledger
ralph ledger load <session-id>
# Search handoffs
ralph handoff search "keyword"
Troubleshooting
Hook not found
# Verify hooks exist
ls -la ~/.claude/hooks/pre-compact-handoff.sh
ls -la ~/.claude/hooks/detect-environment.sh
# If missing, sync from repo
ralph sync-global
Permission denied
# Make hooks executable
chmod +x ~/.claude/hooks/*.sh
chmod +x ~/.claude/scripts/*.py
Context extractor fails
# Test context extractor directly
python3 ~/.claude/scripts/context-extractor.py --project . --pretty
GitHub Repository
Related Skills
sglang
MetaSGLang is a high-performance LLM serving framework that specializes in fast, structured generation for JSON, regex, and agentic workflows using its RadixAttention prefix caching. It delivers significantly faster inference, especially for tasks with repeated prefixes, making it ideal for complex, structured outputs and multi-turn conversations. Choose SGLang over alternatives like vLLM when you need constrained decoding or are building applications with extensive prefix sharing.
evaluating-llms-harness
TestingThis Claude Skill runs the lm-evaluation-harness to benchmark LLMs across 60+ standardized academic tasks like MMLU and GSM8K. It's designed for developers to compare model quality, track training progress, or report academic results. The tool supports various backends including HuggingFace and vLLM models.
langchain
MetaLangChain is a framework for building LLM applications using agents, chains, and RAG pipelines. It supports multiple LLM providers, offers 500+ integrations, and includes features like tool calling and memory management. Use it for rapid prototyping and deploying production systems like chatbots, autonomous agents, and question-answering services.
llamaguard
OtherLlamaGuard is Meta's 7-8B parameter model for moderating LLM inputs and outputs across six safety categories like violence and hate speech. It offers 94-95% accuracy and can be deployed using vLLM, Hugging Face, or Amazon SageMaker. Use this skill to easily integrate content filtering and safety guardrails into your AI applications.
