token-economy
Apply token optimization when writing docs, changelogs, MCP tasks. Quality
When & Why to Use This Skill
The Token Economy skill is a specialized utility designed to optimize text for AI context windows, ensuring maximum information density without sacrificing quality. It streamlines technical documentation, changelogs, and MCP (Model Context Protocol) tasks by applying 15 systematic techniques—such as converting prose to tables and eliminating linguistic fluff. This skill is essential for developers and AI power users looking to reduce token consumption, lower API costs, and improve the performance of LLMs by providing cleaner, more scannable context.
Use Cases
- Technical Documentation Refactoring: Convert verbose software manuals into token-efficient formats using tables and numbered steps to fit more information into the AI's active context.
- MCP Task Optimization: Refine Model Context Protocol task descriptions and progress notes using standardized templates to ensure agents understand requirements with minimal token overhead.
- Changelog and Commit Summarization: Transform lengthy development logs into high-density summaries that preserve critical 'why' context while removing redundant prose.
- Context Window Management: Condense large code samples and repetitive pattern statements into compact references to prevent context overflow in long-running AI chat sessions.
- Cost-Effective Prompt Engineering: Apply 'inverted pyramid' information ordering and conditional verbosity to create prompts that are both effective and economical for high-volume API usage.
| name | token-economy |
|---|---|
| description | Apply token optimization when writing docs, changelogs, MCP tasks. Quality #1, Tokens #2. |
| allowed-tools | Read, Edit, Grep |
Token Economy
Optimize text for AI context: docs, changelogs, MCP tasks, subtasks.
Principle: Quality first. Never sacrifice essential info.
15 Techniques
| # | Technique | Use | Example |
|---|---|---|---|
| 1 | Tables over prose | Comparisons | Function params → table |
| 2 | Bullets with pipes | Multi-part | Part1 | Part2 | Part3 |
| 3 | Numbered steps | Workflows | 1. Step → 2. Step |
| 4 | One example | Code samples | Keep best, remove rest |
| 5 | Pattern statements | Repetition | **Pattern**: All X do Y |
| 6 | Add "Why" | Decisions | **Why**: [reason] |
| 7 | Concrete errors | Debugging | **Error**: \exact message`` |
| 8 | No fluff | Decorations | Remove ASCII art, emojis |
| 9 | Scannable | Structure | Headers, bullets, tables |
| 10 | Consolidate | Duplicates | Merge similar sections |
| 11 | Compact code | Examples | 3-5 key lines only |
| 12 | Quick lists | Reference | Command/param tables |
| 13 | Inverted pyramid | Info order | Critical → Details → Context |
| 14 | Conditional verbosity | Complexity | High for complex, low for simple |
| 15 | No teaching | Reference | State facts, skip basics |
Quick Workflow
- Identify: Prose → tables? Multiple examples → one? ASCII → numbered? Teaching → facts?
- Apply: Convert, consolidate, compact
- Check: Essential info preserved? Scannable? No fluff?
MCP Templates
Task:
Requirements: [WHAT] | Files: [PATH:LINES] (action) | Acceptance: [CRITERIA] | Why: [CONTEXT]
progress_notes: [ACTION] [WHAT]. [NEXT]
completion_summary: [DONE]. [TECH]. Files: [PATHS]. [IMPACT]
Examples
Before (prose):
This function validates user input. Takes string input_text, returns bool.
If valid returns true, otherwise false. Checks empty, special chars, length ≤100.
After (compact):
**validate_input(input_text: str) → bool**: True if valid (non-empty, no special chars, ≤100)
Before (ASCII art - 180 tokens):
┌─────────┐
│ Auth? │
└────┬────┘
▼
┌─────────┐
│ Valid? │
└──┬───┬──┘
Y N
After (numbered - 36 tokens):
1. Check auth
2. Valid → YES: Process | NO: Reject
Common Fixes
| Issue | Fix |
|---|---|
| Multiple examples | Keep 1 (#4) |
| Prose comparison | Table (#1) |
| Teaching | Facts only (#15) |
| ASCII art | Numbered (#3) |
| No "why" | Add context (#6) |
See TEMPLATES.md for copy-paste templates.