AI 编码Agent上下文效率优化检查清单

You are an expert AI coding agent optimization consultant. Generate a comprehensive efficiency checklist for AI coding agents (Claude Code, Cursor, Codex, Maki, etc.).

The checklist should cover:

1. Context Window Optimization

   • File indexing strategies: tree-sitter skeletons vs full reads
   • When to use code_execution/sandbox for data filtering vs passing raw output
   • Subagent delegation: weak/medium/strong model selection criteria
   • System prompt compression techniques

2. Session Memory Management

   • CLAUDE.md / .cursorrules / AGENTS.md best practices
   • Session summary and context injection patterns
   • When to start fresh vs continue sessions

3. Token Budget Tracking

   • Estimating cost per task type: bug fix, feature, refactor
   • Read vs write token ratio optimization
   • Bash output filtering and truncation strategies

4. Task Decomposition

   • Breaking large tasks into parallelizable subtasks
   • Plan-review-execute workflow
   • When to use background agents vs inline execution

5. Quality Gates

   • Pre-commit validation checklist
   • Test coverage requirements before PR
   • Diff review before apply

For each item, provide:

• Best practice
• Common mistake
• Expected token savings or quality improvement

Output as a structured markdown checklist ready to paste into a project's AGENTS.md or CLAUDE.md.