Token Saver Master

Problem

In long conversations, multi-turn tasks, and large-document workflows, AI answers can become bloated: repeated background, over-explained steps, and paragraph-heavy output consume context and tokens while diluting the key constraints. Token Saver Master targets that kind of expressive redundancy and context bloat, not raw model capability.

How It Works

The skill turns compression into concrete editing actions:
- Input compression: remove redundant wording, extract core instructions, constraints, and relevant context, and reduce repeated descriptions.
- Structured reasoning: use an outline such as problem → solution → action, with symbols like →, ∴, √, and × to reduce verbose intermediate explanation.
- Concise output: prefer conclusions first, lists, and tables over long paragraphs, while keeping the necessary evidence.
- Context management: mark key decision points, index concepts with keywords, summarize history, and limit stale context from entering later turns.
- Code handling: preserve code blocks exactly, while tightening comments, blank lines, and docstrings where that does not hurt readability or runtime correctness.

It defaults to balanced mode for day-to-day queries and engineering discussion. Triggers include /省token, /token on, /token off, or keywords such as summarize and condense. The claimed total saving range is 30-50%, estimated across input, reasoning, output, and context dimensions; actual savings vary by task complexity and model style.

Boundaries

It fits long chats, complex reasoning, code generation, large-document cleanup, and multi-turn tasks. When users ask for more detail, when code tasks are complex, when follow-up questions require evidence, or when comprehension is difficult, the skill can automatically switch to detailed mode or restore fuller output. For formal clauses, legal-style wording, or step-by-step justification, prefer detailed mode over the shortest possible answer.