省Token助手：在 WorkBuddy/QClaw 里把对话 Token 花得更值
省Token助手（@user_ae5129cb/shengtoken-zhushou，v1.0.0）是 SkillHub 社区 ai-agent 类 Skill，作者青木会江湖，无需 API Key。提供任务模板、优化提问、上下文管理与批量创作指导，面向 WorkBuddy/QClaw 用户。本文说明定位与四项能力，并依据 skillhub.cn 安装文档给出 CLI 安装命令及 Cursor、Claude Code、Codex、workbuddy 等 skills 目录要点。
Read Moretoken秘书就一个字省token：把本地检索与压缩接到 Agent 流程里
SkillHub 社区 Skill「token秘书就一个字省token」（@user_5b2fc4e3/token-sheng，v1.0.0）将本地知识检索、情报压缩与云端精加工写入 SKILL.md，配合 GLOBAL_RULES Agentic RAG 以缩短上云上下文。文章说明触发方式、五步流程、skillhub 安装命令与 Cursor 等 skills 目录，并提示包内主要为流程约定，脚本与知识库需自行部署；README 另述 Ollama 长文压缩与 config 示例。
Read MoreToken省流skill：用执行模式路由控制 Agent 的 Token 消耗
Token省流skill（@user_6fc816d6/token-efficient-task-router-pro-version，v2.8.0）是 SkillHub 社区发布的控制层 Skill，面向 WorkBuddy、OpenClaw、QClaw 等 Agent：通过意图识别将请求路由到 Ask、Plan、Craft、Expert，配合预算先行、Token 监控与预估、消耗策略、上下文快照压缩与模型能力画像路由，并以分阶段确认减少未授权的大范围编辑。本文基于 SkillHub 公开简介与安装文档，说明定位、能力、skillhub CLI 安装路径及适用注意点。
Read Moretoken节省大师：面向 OpenClaw 链路的 Token 优化 Skill
token节省大师是 SkillHub 社区 Skill（@user_65cfe060/token-saver-king，v1.0.0），面向 OpenClaw 链路，宣称通过智能压缩、上下文管理、自动摘要与代码优化等多引擎降低 Token 消耗。本文依据 SkillHub 页面与安装文档说明定位、能力与 skillhub CLI 安装步骤（含 Cursor 等 skills 目录），并提醒节省比例需自行验证、摘要可能影响原文精度。无需 API Key（requires_api_key=false）。
Read More古代人：用超压缩中文回复降低 Agent 对话 Token 消耗
古代人是 SkillHub 社区 Skill（@user_3a8d4fb8/ancientman-cn，v1.0.1），面向中文大模型提供超压缩通信模式，简介称可减少约 75% 令牌使用并保持技术准确，无需 API Key。支持豆包、DeepSeek、千问等适配，并提供古风小生文言文压缩子模式。文章说明定位、触发词、SkillHub CLI 安装路径及 Cursor/Claude Code 等 skills 目录注意事项，便于开发者评估是否用于长对话降本。
Read MoreToken节省王：在 Agent 全流程中压缩 token 消耗的 Skill
Token节省王（@user_65cfe060/burgess-123）是 SkillHub 社区 ai-agent 类 Skill，面向 OpenClaw 问题解决全流程，通过输入压缩、思考精炼、输出瘦身与上下文管理降低 token 消耗，支持 minimal、balanced、detailed 自动切换。官方页面宣称整体约可省 20%–40%，无需 API Key。本文基于 SkillHub 资料说明四引擎、安装命令（skillhub install 与 Cursor/Codex 等 skills 目录）及长对话与代码生成等适用场景与使用注意。
Read More词元token节约省钱大师：在 Agent Skills 里做 Token 诊断与瘦身
词元token节约省钱大师（@user_ad872d55/token-saver-master，v2.2.2）是 SkillHub 社区 ai-agent 类技能，用于诊断 Skills 冗余并执行压缩、去重与删除以释放 token，且要求不得为省钱削减数据、步骤与来源等干货。本文基于 SkillHub 公开说明整理定位、红线、skillhub 安装路径（含 Cursor 的 ~/.cursor/skills/）及与官方一致的触发用语，便于在维护膨胀技能库时可控瘦身。
Read Moretoken省流助手：在 Agent 对话前压缩上下文，节省 Token
SkillHub 社区 Skill「token省流助手」（@user_29dc410e/token-saver-qclaw，v1.0.1，ai-agent）面向长对话与大段代码粘贴场景，自动压缩用户输入、代码片段与对话内容，官方称可节省约 20%–70% Token，无需 API Key。文章说明其定位、已核实功能，以及通过 skillhub CLI 或 AI 提示词安装到 Cursor、Claude Code 等 skills 目录的步骤与使用注意。
Read More记忆管家 V3.9.0：管理 `.workbuddy/memory/` 的 Token 节流与相关性加载
记忆管家 V3.9.0 是 SkillHub 上的 AI 记忆加速器与 Token 节流阀，面向 .workbuddy/memory/ 目录。核心包括 tiktoken 分词计量、--query/--tags 相关性加载（CJK bigram，可选 embedding）、硬预算闸与 throttle 压实、三色告警与近 7 日均值、--force 与全命令 --json、self-test 安全自检，以及并行摘要与 rank 缓存等性能优化。文章说明安装 skillhub CLI、--dir 指向各 Agent skills 目录的方式，以及 Token 检查、智能加载、清理备份等典型触发场景与使用注意。
Read Morecontext-compression: Context Compression and Token Optimization for Long-Session Agents
context-compression, from Agent-Skills-for-Context-Engineering, focuses on context compression and handoff summaries for ultra-long conversations. The article sets the optimization target as tokens-per-task, introduces three paths—anchored iterative summarization, opaque compression, and full-summary regeneration—emphasizes artifact trails and mandatory sectioned summaries, and provides trigger strategies, probe-based evaluation methods, and common enablement pathways such as Claude Code and npx skills.
Read Moredsh-token-frugal: DeepSeek Harness Input Token Cost Optimization Plugin
The DeepSeek Harness plugin dsh-token-frugal is designed to reduce input token costs for long sessions through tiered saving strategies and session memory matching. It extends five key areas: lossless or truncated compression of tool results, agent tool directory management, workspace memory matching, a visual tiered control panel, and a statistics page. Users can adjust the saving level in real time below the chat box and view detailed token-saving statistics in settings, effectively controlling costs without compromising core capabilities.
Read Moredsh-pwsh-rtk-rewrite: A Native rtk Command Bridging Plugin for DeepSeek Harness
`dsh-pwsh-rtk-rewrite` is a native rtk bridge plugin for DeepSeek Harness (DSH), designed to use the rtk rewrite mechanism to convert native shell commands into compact rtk commands and save tokens. It supports Windows + PowerShell and provides transparent rewriting, preservation of sandbox semantics, and automatic fallback on failure. By automatically segmenting and rewriting composite commands (e.g., `git status` -> `rtk git status`), it addresses verbose native shell output while keeping the tool-call logic unchanged.
Read Moredsh-token-saver: DeepSeek Harness Plugin
`dsh-token-saver` is a Cordis plugin designed for DeepSeek Harness, optimized to reduce token consumption in long-running tasks by mounting extension points. Core features include lossless head-and-tail trimming, grep result folding, automatic context compression, and optional LLM-based summarization. It offers conservative, balanced, and aggressive modes. Benchmarks show the balanced mode saves approximately 98.5% of tokens on average. Installation requires Node.js >= 22.19.0, configuration is flexible, and it is well suited to scenarios with large files and frequent retrieval results.
Read MoreDeepSeek Harness Plugin: merenguesl/dsh-tool-normalizer
`dsh-tool-normalizer` is a self-healing plugin for DeepSeek Harness (DSH), located in the `tools/execute` pipeline, designed to fix failed model tool calls. By filling in missing `description` fields, resolving relative paths, automatically repairing schemas, and retrying file operations, this plugin reduces full-token retries and billing consumption caused by errors, significantly improving the stability and reliability of the agent.
Read Moredsh-context-guard: A Token-Efficient Tool Result Pruning Plugin for DeepSeek Harness
`dsh-context-guard` is a plugin for DeepSeek Harness (DSH), intended to address context overflow caused by the large size of tool results in long sessions. It controls token consumption by pruning oversized tool results. Specifically, it replaces regular results with an approximately 2.3K-character preview (retaining the beginning and end) before the next request, while storing the original data in an append-based history for targeted recovery. The plugin supports search and recall strategies and must be used with specific threshold settings and a Persona prompt. DSH must be restarted after installation.
Read Moredsh-plugin-context-manager: A DSH plugin that automatically records conversations, injects memory, and folds real context.
# Introduction Running long sessions in DeepSeek Harness (DSH) often presents two common challenges: First, the limited context window means that once early conclusions are squeezed out, the model "forgets" them; Second, when switching to a different session, previously recorded key points cannot be carried over. The common solution is to inject summaries into the system prompt, but this only affects auxiliary prompt blocks, while the real dialogue history continues to grow, leading to increasing token consumption. The `dsh-plugin-context-manager` introduced below addresses these two issues: it automatically records every round of dialogue, injects memory and prioritized cross-session notes into each model request, and directly collapses the real dialogue history visible to the model to reduce token usage. # What is dsh-plugin-context-manager
Read Moredsh-mcp-lazy: DSH's Lazy-Loading MCP Plugin, Reducing Token Waste by Disclosing Tool Schemas on Demand
# Introduction The philosophy of DSH (DeepSeek Harness) is "everything is a plugin." Developers often install multiple MCPs on it, such as files, browsers, and databases. This incurs a specific cost: even if the current problem only requires file tools, all MCP tool names, descriptions, and parameters may enter the model context simultaneously. The more MCPs installed, the more tokens are consumed per round. Below is an introduction to leaforbook/dsh-mcp-lazy, a DSH plugin designed to hide descriptions of temporarily unused tools and load them when the task requires them. # What is it dsh-mcp-lazy (npm package name `@yilinxiao/dsh-mcp-lazy`) is an MCP for DeepSeek Harness (DSH).
Read MoreDeepTrial/dsh-bash-rtk: Pass DSH bash commands meeting conditions to rtk for compressed output
`dsh-bash-rtk` is a DSH bash executor plugin maintained by DeepTrial, designed to route eligible Shell commands to `rtk` for processing to compress tool output and reduce Token consumption while preserving the original execution semantics. The plugin adopts a conservative approach, only applying `rtk` wrapping to simple commands within a whitelist (such as `git status`, `cargo build`). For complex commands containing Shell metacharacters, non-whitelisted commands, or when `rtk` is unavailable, it automatically falls back to pass-through execution. It provides both standard and sandboxed Executors and supports forced activation via configuration settings. Prior usage requires separate installation of `rtk`, and the plugin is disabled by default, requiring explicit enablement in the configuration to take effect.
Read MorePonytail: Let AI Agents Follow the YAGNI Principle to Reduce Over-Engineering and Token Waste
# Ponytail: August 2026 GitHub Trending Project Ponytail is a YAGNI (You Aren't Gonna Need It) agent skill suite designed specifically for coding agents. Through decision ladders and commands such as `/ponytail-review` and `/ponytail-audit`, it prioritizes reuse, standard libraries and native capabilities before writing code, curbing over-engineering and unnecessary dependencies. Official agentic benchmarks conducted on real FastAPI+React repositories show an average code reduction of approximately 54%, a token reduction of around 22%, a cost reduction of roughly 20%, while maintaining 100% security. It supports over 14 hosting environments including Claude Code Plugin and Cursor rule injection. This article introduces its principles,实测 data, installation methods and applicable boundaries.
Read MoreMCP Server codebase-memory-mcp: Enabling AI Agents to Truly "Remember" Your Codebase
In July 2026, the GitHub Trending hot project codebase-memory-mcp (produced by DeusData) indexes codebases into persistent knowledge graphs via the MCP protocol. It supports parsing 158 programming languages and provides 15 MCP tools, enabling agents such as Claude Code and Cursor to replace file-by-file exploration with structured queries. According to official benchmark tests, the token consumption for structural queries can be reduced by about 120 times, and indexing for Linux kernel-level repositories can be completed in approximately 3 minutes. This article introduces its Hybrid LSP semantic enhancement, core tools, installation method and applicable scenarios.
Read More