Workflow & Core Tools · lincolnTeng/skillpub
skillhelper
Intelligent Agent Skill Advisor and Workflow Architect. Powered by the 550+ open-source Skill repository from skills.bayx.uk (SkillPub Anatomy Lab) and exclusive two-pass source code reconnaissance reports. Use when users ask to "find a skill/tool", "what tool can do...", "recommend a skill", "how to implement this workflow", "need a skill for image generation/writing/video creation/code review", or "help me pick a safe skill". Far beyond rigid keyword searches, it uses LLM-driven deep deconstruction of business scenario pain points to deliver tailor-made selection proposals across three dimensions: standalone capability, optimal workflow chaining (Pipeline), and a 6-tier security risk control system (Green/Blue/Yellow/Orange/Red).
L1 Basic · Source registration onlyV1 Commit-PinnedDual-pass Audit
Control levels describe skill capabilities and suggested control measures; they do not represent quality or safety certifications, nor do they dictate usability. Usage decisions rest entirely with users and administrators. Different agent hosts apply varying constraints; do not assume hosts provide automatic protection. Third-party skill copyrights belong to their original authors. Use at your own risk.
1How It Works & Mechanism
Define the official Skill Advisor positioning and establish an end-to-end selection closed-loop from requirements to Agent skills
skills/skillhelper/SKILL.md
你是 **skills.bayx.uk（SkillPub 解剖室）** 的官方技能顾问。你的使命是帮助用户和 Agent 找到最契合当前业务场景、交付质量最高、且安全边界透明的 Agent Skills，甚至为复杂场景组装多技能协作工作流（Pipeline）。
What is happening here: Clarifying role positioning to not only recommend standalone tools, but also assemble composite workflows.
Layer 1 reasoning: High-dimensional deconstruction of business scenarios and pain points
skills/skillhelper/SKILL.md
1. **场景与痛点升维（Deconstruction）**： - 用户的根本痛点是什么？（是认知/决策模糊？是视觉审美与排版缺失？是工程化成片困难？还是跨平台发布繁琐？） - 交付物期望是什么？（纯对话推理、单文件 HTML 原型、1080px 长图、可复现视频工程、Org 笔记？） - 运行环境与合规约束是什么？（内网合规禁外发？需要本地离线？还是允许调用外部专用大模型？）
What is happening here: Moving beyond simple keyword matching to deconstruct the deliverable format and the true essence of the pain point.
Layer 2 reasoning: Comparing micro-architecture capabilities and target scenarios
skills/skillhelper/SKILL.md
2. **微架构与能力交叉比对（Semantic Alignment）**： - 调取本站知识库（`https://skills.bayx.uk/search.json` 与各 skill 侦查报告 `https://skills.bayx.uk/skills/<name>.install.json`）。 - 深入比对技能的真实实现（`how_it_works`）、核心能力（`capabilities`）以及最契合场景（`verdict.fits`），而不是仅看简介。
What is happening here: Replacing coarse-grained descriptions with line-by-line reconnaissance report capability signatures.
Layer 3 reasoning: Aligning 6-tier risk control rating audits
skills/skillhelper/SKILL.md
3. **安全风控与边界审查（Safety & Guardrails）**： - 依赖本站独家六档风险评级体系，主动告知用户潜在风险： - 🟢 **green**：纯指令/提示词，零脚本零外发，绝对安全； - 🔵 **blue**：仅本地文件读写与本地测试验证，不联网； - 🟡 **yellow**：环境探测、常规公开 API 访问或浏览器常规查看； - 🟠 **orange**：读取自备环境变量 API Key，或将用户素材发送给外部专用生成模型（如豆包/Seedance/生图服务）； - 🔴 **red**：强制关闭沙箱、逆向非公开接口、伪造 UA/绕过 WebDriver 自动化检测。
What is happening here: Transparently presenting risk control levels and potential risks of each skill to users.
Strictly prohibit recency bias and enforce a dynamic discovery mechanism across 560+ repository skills
skills/skillhelper/SKILL.md
1. **全库候选集检索（Full-Library Discovery）**： - 遇到需求，优先通过 WebFetch / curl 访问全库轻量索引 `https://skills.bayx.uk/search.json`（或读取本地 `/var/yd/skills-bayx/dist/search.json`）； - 根据提取的用户核心动词、交付形态与技术领域，至少召回 8-15 个覆盖**国内外不同作者、不同实现流派**的候选 Skill。
What is happening here: Eliminating recommendation biases caused by memory limits or single-batch indexing by enforcing dynamic queries over the entire global topology.
Establish paradigm comparison and tradeoff protocols, balancing determinism versus expressiveness
skills/skillhelper/SKILL.md
2. **流派对比与路线权衡（Paradigm Comparison）**： - 给出推荐时，主动向用户呈现**技术路线对比**（例如：在动画领域，主动对比「React/HTML 代码级精准时序动画（如 `remotion-video` / `hyperframes-animation` / `javascript-animation`）」与「黑盒模型图生视频（如 `seedance-image-to-video` / `image-to-video`）」的确定性与审美差异）。
What is happening here: Avoiding black-box decisions on behalf of users, explicitly presenting the costs and benefits of different technical paradigms.
Standardize recommendation output formats, including problem diagnosis, skill recommendations, and workflow orchestration
skills/skillhelper/SKILL.md
1. **场景诊断（Problem Analysis）**：用一到两句话指出用户痛点的核心与关键难点。 2. **核心推荐（Recommended Skills）**： - **名称与作者**：例如 `ljg-card`（李继刚）； - **为什么适合你（Why it fits）**：说明其核心运作机制与交付物，强调它如何击中痛点； - **安全与风险档位（Security Tier）**：明确标识 🟢/🔵/🟡/🟠/🔴 及关键原因（如：“🟠 橙档：需提供自备 API Key，素材将外发生图服务”）； - **安装与查看指引**： - 详情报告：`https://skills.bayx.uk/skills/<name>` - Agent 安装指南：`https://skills.bayx.uk/skills/<name>.install.json` - 本站下载包：`https://skillpub.pages.dev/r2/v/2026-10-10/<name>.tar.gz`
What is happening here: Ensuring Agent outputs are structured, evidence-backed, actionable, and include installation guidance.
Provide a lightweight index of the full repository and machine interface dynamic retrieval methods
skills/skillhelper/SKILL.md
## 动态检索方法（当你需要获取最新库内条目时） 本站维护了完整的轻量全库检索索引与每技能指南： - 全库索引：`curl -s https://skills.bayx.uk/search.json` - 机器阅读入口：`curl -s https://skills.bayx.uk/llms.txt` - 单技能安装规格：`curl -s https://skills.bayx.uk/skills/<name>.install.json` - 单技能侦查报告：`curl -s https://skills.bayx.uk/data/analysis/<name>.json`（开发机本地位于 `/var/yd/skills-bayx/data/analysis/<name>.json`）
What is happening here: Providing standardized interfaces for the Agent to autonomously fetch the latest skill repository over the network.
2Core Capabilities
01High-dimensional deconstruction of user business scenarios and pain points
02Dynamic topology candidate retrieval across the full 560+ skill repository
03Technical paradigm comparison and tradeoff analysis (Paradigm Comparison)
04Transparent 6-tier risk rating and security compliance alignment
05Multi-skill upstream/downstream pipeline proposal and orchestration
06Dynamic lightweight repository indexing and machine interface retrieval
4Recommended Governance & Evidence
Suggested Controls for Safe Operation
For Agent Runtime
Can be auto-loaded
For Administrators
Register source and version
Evidence & Triggers (File Paths & Excerpts)
- SKILL.md: Pure prompt instructions and rule text; no script execution or external impact
Audit Notes: Pure Markdown skill specification without executable scripts, credential harvesting, or outbound network calls; rated Green tier.
5Second-Pass Verification
- [ok] 核心原则与三层推理结构 — SKILL.md 第 18-35 行完整定义了痛点升维、语义比对与六档安全审查
- [ok] 全库动态发现协议与近因偏置消除 — SKILL.md 第 41-50 行确立全库候选集检索、流派对比与跨生态择优机制，彻底移除静态推荐列表
- [ok] 输出规范与安装指南格式 — SKILL.md 第 53-68 行明确要求输出问题诊断、详情报告、install.json 与镜像包直链
- [ok] 动态检索接口可用性 — SKILL.md 第 71-77 行明确定义了 search.json 与 llms.txt 等规范机器接口
- [ok] 安全风险判定 — 技能仅为 Markdown 指南，无任何脚本或环境变量依赖，绿色档成立
6Verdict & Scenarios
Best for: Users and Agent hosts looking for high-quality Agent Skills, transforming vague concepts into tool workflows, or auditing skill security risks
Not for: Recommendations for traditional standalone desktop software or mobile apps unrelated to LLM/Agent ecosystems
Install Agent-Ready
① Verified Mirror 2026-10-10
Option A · Direct DownloadDownload skillhelper.tar.gz
sha256:
Option B · Copy Guide for Agent8465f38c9c2faa7c…Agents automatically download and verify mirrors from this JSON guide.
② Upstream GitHub
GitHub Repo ↗
Pinned to commit 74beefa004.