360 Secure SkillHub
360 安全技能中心

Token Optimization

核心功能是基于已落地的规范流程，在保障回答质量的前提下削减不必要的令牌消耗，仅采用已实现的优化规则，不涉及实验性策略。典型场景含优化令牌消耗、精简上下文冗余、规范回复、对齐多助手用牌行为。输入为用户相关需求，输出为符合现有规范的当前优化方案，更新需版本化规范。

sinoslug
sinoslug
热度 37.77
最近更新 2026-05-11 17:40:23
分类：效率依赖：读写文件应用场景：摘要总结 / 工作流

安装技能

复制下面的话术发给你的 Agent，它会通过 secure-skills 检索并安装该技能。

执行命令安装此 skill： npx secure-skills-cli add token-optimization-sinoslug
Metadata
name
token-optimization
description
Reduce unnecessary token usage with the current production-proven workflow. Use when users ask to optimize token consumption, reduce prompt/context bloat, standardize concise replies, or align token-saving behavior across assistants. Includes only currently implemented practices (no future/experimental tactics).

Token Optimization

Use this skill when the task is about reducing token usage while keeping answer quality stable.

Workflow

  1. Read the current baseline spec first:
    • references/current-spec.md
  2. Apply only the already-implemented rules from that file.
  3. Do not introduce speculative or not-yet-implemented optimization tactics.
  4. If asked to change behavior, update the spec and version it.

Scope guard

  • In scope: concise reply behavior, log/tool-output trimming, current token-show usage flow, and currently active execution process.
  • Out of scope: future ideas not yet implemented (auto-threshold switching, automatic split replies, weekly auto-reviews), unless user explicitly requests a new version update.

Output expectation

When asked to explain or sync the optimization plan, provide the “current-state only” version and cite that it follows references/current-spec.md.