Prune Context

Context Pruner

When to activate

Activate when the user runs the /prune-context command, or when session token consumption is high, or when the context window is getting bloated in a long session.

When NOT to use

Do not use at the beginning of a session, or for simple queries that do not require state persistence or context reduction.

Instructions

1. Pause current tasks and review the conversation history of the active session.
2. Summarize the session into a highly condensed, token-efficient state:
   • Active Objective: The current high-level goal you are working on.
   • Completed in Session: A concise bulleted list of completed changes, tests, and files.
   • Key Architectural Context: Crucial project details, APIs, or files discovered during the session.
   • Next Steps: Explicit checklist of what needs to be done next.
3. Write this summary to .claude/pruned_context.md.
4. Output the summary in the chat so the user can see it.
5. Instruct the user to run the /clear command to reset the context window and wipe the token bloat.
6. Provide the command or text they should input after clearing to resume: "Please run /clear now to reset memory. Once cleared, prompt me with: 'Resume from .claude/pruned_context.md' to restore this state."

Example

User: /prune-context Claude: Starting context pruning to reduce token bloat. [Claude writes summary to .claude/pruned_context.md]

Current Session Snapshot

• Active Objective: Restructure README documentation and add fail-fast enforcer rules.
• Completed: Created rules/common/fail-fast.md and hooks/post-tool-use/fail-fast.sh.
• Key Context: The compiler is run via npm test and require specific git authoring config.
• Next Steps: Implement Plan-First pre-tool-use hook.

State saved to .claude/pruned_context.md. Please run /clear now to reset memory and clear token bloat. Once cleared, prompt me with: Resume from .claude/pruned_context.md to restore this state.