General rules

Optimize for token efficiency

• Use sub-agents to process documents (code, configuration, or other content) required for the task.
• Prefer programmatic extraction and classification over reading full document contents.
• Store hashes to detect changes programmatically, enabling caching of prior results.
• Generate and reuse utility scripts to avoid duplicate AI inference.

Don't reinvent the wheel

The domain may already have established programmatic ways to map system state or generate a documentation index. If such solutions exist, evaluate whether they fit your situation. Building on an existing, well-tested domain-specific solution is usually better than rolling your own, unless it requires significant changes or has unacceptable limitations.

Preparation

Establish detailed domain-specific documentation before working on actual tasks to improve token efficiency down the line.