Antigravity Context Tree-Shaking Skill (/shake)

Compact conversation context in-place, freeing 50%–80% of transmitted tokens while preserving 100% of user prompts, assistant reasoning, thoughts, error tracebacks, and the last 10 user conversational turns in active working memory.

---

⚡ Execution Workflow

1. Identify Conversation Context Paths

From your runtime context:

• Conversation ID: <conversation-id>
• Transcript Path:

  <appDataDir>/brain/<conversation-id>/.system_generated/logs/transcript.jsonl
  

• Artifact Directory:

  <appDataDir>/brain/<conversation-id>/
  

---

2. Execute High-Performance Pruner

Run the native binary using run_command:

# Direct in-place compaction, line-indexed receipts, and interactive artifact generation
~/.gemini/bin/shake-prune "<appDataDir>/brain/<conversation-id>/.system_generated/logs/transcript.jsonl" "<appDataDir>/brain/<conversation-id>/"

Pruned outputs include exact line=N pointers to the backup file. If you ever need to inspect historical commands or file contents from earlier turns, invoke view_file(archive, StartLine: line, EndLine: line).

---

3. Display the Report & Continue

The binary outputs a formatted markdown summary. Present it directly to the user:

• Report the tokens saved and physical reduction percentage.
• The PreInvocation hook pins the assistant's working focus to the clean state.
• The user can immediately continue typing prompts in the exact same chat window.