Token-Efficient Reading

# Token-Efficient Reading

AI Summary

Purpose:

• Defines how agents should avoid wasting tokens while using the repository.

Key points:

• Agents must read indexes and summaries before full documents.
• Every long file should have frontmatter and AI Summary.
• Full body reads require a clear reason.
• Directory indexes act as navigation maps.

Relevant when:

• Any AI agent starts work.
• Creating or updating Markdown files.
• Debugging context overuse.

Do not read full document unless:

• You need detailed reading conditions.

Linked documents:

• AGENTS.md
• ai/index.md
• ai/wiki/index.md

Reading Order

1. AGENTS.md
2. ai/index.md
3. relevant directory index
4. target file AI Summary
5. full body only if necessary

Full Body Conditions

Read the full body only if:

• the user explicitly requests full review;
• the summary is missing or insufficient;
• exact details are required;
• the document is the primary source;
• summary and body conflict.