token-optimizer

name	token-optimizer
description	Optimizes token usage and context efficiency for the agent. Use when performing complex tasks to minimize costs and prevent context bloat.

Token Optimizer

Expert guidance for maximizing token efficiency. High-performance, low-cost operation.

Mandates

• Surgical Actions: Never read whole files when lines suffice. Use grep_search to target before read_file.
• Parallelism: Group independent tool calls in a single turn.
• Zero Filler: Eliminate all conversational fluff. Every word must carry technical signal.
• Context Pruning: Avoid re-reading recently accessed files. Rely on short-term memory.
• Sub-Agent Delegation: Delegate high-volume or repetitive tasks to sub-agents to keep the main history lean.

Workflow

1. Analyze Request: Identify minimal data needed.
2. Surgical Research:
   • Use glob + grep_search (parallel) to find symbols.
   • read_file with start_line / end_line ONLY.
3. Compressed Execution:
   • Batch edits to different files in one turn.
   • Use replace with minimal context while ensuring uniqueness.
4. Verification:
   • Run targeted tests/commands. Capture ONLY relevant output (tail logs).

Compression Level: Ultra

• Abbreviate prose.
• Use symbols (->, =, !, ?).
• Drop articles and conjunctions.