Publish To All Skill Stores

io.github.xixi048由 GitHub 帳號 xixi048 發布v2.0.1更新於 Oct 26, 2026

List, check and package a SKILL.md skill for every agent-skill store.

發布方自己的服務:Publish.fun

已驗證STDIO僅桌面效率與工作流開發者工具

概覽

AI 產生的概覽

本機 MCP 伺服器,協助助理列出各類 agent skill 商店、對 SKILL.md 技能做就緒檢查,並產生各商店要求的打包格式。

可以做什麼
  • 查詢 37 個 agent skill 商店的登錄表,含地區、定價、提交形式與驗證等級
  • 對技能資料夾做提交前就緒檢查(目錄結構、授權、危險呼叫、網路)
  • 依不同商店要求產生 zip 與目錄結構
  • 依商店類型提供提交指引:GitHub 自動索引、公開提交 API、GitHub PR、需登入的 CLI 或 OAuth 表單、平台託管
適用情境
當你已有 SKILL.md 技能,想同時上架到多個 agent skill 市集、又不想逐一重新摸索提交規則時使用;也可在提交前檢查技能是否就緒。
執行需求
以本機 stdio 程序執行,需 Python 3.8+(僅標準函式庫),README 範例直接啟動 mcp/server.py。未宣告任何帳號、API 金鑰或環境變數。僅在主動提交時存取所選的公開端點。
安裝前請注意
批次提交腳本會代你向公開商店端點送出提交,請先確認送出內容與目標頻道。需要登入或 OAuth 的商店須在腳本之外處理。隨附文件涉及付費頻道與分潤,付費前請先確認條款。

安裝

在 SourceWeft 中

  1. 開啟 儀表板中的 Publish To All Skill Stores,將其新增到工作區。
  2. 為需要使用其工具的對話啟用該服務。

Desktop only,透過 STDIO。 STDIO 服務會啟動本機處理程序,因此需要 SourceWeft 桌面主機。

其他 MCP 客戶端

參照 儲存庫 中的啟動說明。

README

publish-to-all-skill-stores

[License: MIT]

同时上架所有 Skill 商店 · an agent skill for pushing an existing SKILL.md skill to agent-skill marketplaces — at home and abroad, free and paid — without wasting a week rediscovering how each one actually accepts a submission.

中文说明见下。

What it is

A playbook, not a wrapper. It tells an agent how each class of store actually works, so no time is wasted on the ones that will never accept a file drop:

FormSubmission actionExamples
Auto-indexed from GitHubfix the repo (path + topics + README)skills.sh, SkillsMP, claude-plugins.dev
Public submit APIone HTTP POSTskillstore.io, skillsmd.dev, skillbay.sh
GitHub PRfork → add file → PRbuildwithclaude, awesome lists
Login-gated CLI / OAuth formlog in firstClawHub, Tank, Cow Skill Hub, agentskill.sh
Platform-hostedport the skill — a different projectCoze, AgentMore, GPT Store

What's new in v2

  • channel registry 24 → 37 (adds LobeHub, agentskills.in, XFYUN Astron, the MCP-side directories, official repos, the CN second tier, XiaPing, and the /.well-known/agent-skills/index.json discovery endpoint)
  • a named rule for whether a store is real: no receipt id and no PR means it did not happen
  • paid research: payment depends on form, not on channel — Tencent SkillPay is 0-cut but only accepts backend-backed "Pay Skill"

Install

bash
npx skills add xixi048/publish-to-all-skill-stores
text
/plugin marketplace add xixi048/publish-to-all-skill-stores

What's inside

skills/publish-to-all-skill-stores/├── SKILL.md                       the decision-first workflow├── references/│   ├── store-registry.md          24 channels: region, pricing, form, verification level│   ├── submission-recipes.md      6 recipes + SPA recon + mirror/token push│   ├── paid-channels.md           revenue splits, skillbay field table, price tiers│   └── pitfalls.md                20+ "symptom → cause → fix" entries└── scripts/    ├── preflight.py               readiness check (layout, license, danger calls, network)    ├── build_package.py           build the shapes each store wants    └── submit_all.py              batch submitter (--registry for the full table)

Quick start

bash
# 1. readiness check — ERROR means don't submit yetpython skills/publish-to-all-skill-stores/scripts/preflight.py <repo>/skills/<name>
# 2. build the package shapespython skills/publish-to-all-skill-stores/scripts/build_package.py <repo> skills/<name> --out dist
# 3. see every channel and its submission formpython skills/publish-to-all-skill-stores/scripts/submit_all.py --registry
# 4. batch submit the ones that need no loginpython skills/publish-to-all-skill-stores/scripts/submit_all.py --all --repo <github-url>

The one rule that pays for itself

Directory sites only index skills/<name>/SKILL.md. A skill sitting at the repo root is never discovered and never errors — it simply stays invisible. Move it, fix the relative paths, re-submit.

Use as an MCP server

The same three tools are exposed over MCP (stdio, zero dependencies), so any MCP-capable agent can call them directly:

json
{  "mcpServers": {    "publish-to-all-skill-stores": {      "command": "python",      "args": ["C:/path/to/publish-to-all-skill-stores/mcp/server.py"]    }  }}

Tools: skill_store_registry (37 channels), skill_preflight (readiness check), skill_build_package (zip shape stores ask for).

Machine-discoverable index

docs/.well-known/agent-skills/index.json is published on GitHub Pages so agents that honour the /.well-known/agent-skills/index.json convention can find these skills without any store:

text
https://xixi048.github.io/publish-to-all-skill-stores/.well-known/agent-skills/index.json

Requirements

Python 3.8+, standard library only. No network calls from the scripts except the public endpoints you ask for.

License

MIT — see LICENSE.

來源:README.md,提交 078ab79

工具

0
SourceWeft 曾在隔離沙箱中以無憑證方式啟動該軟體包一次,以列出其工具,但未能完成:該軟體包類型無法在沙箱中執行。

版本歷史

1
  1. v2.0.1最新Oct 26, 2026

評分

已滿足適用檢查項的 50%

該評分衡量此收錄條目的完整程度,依據註冊中心條目、端點與公開來源計算得出。它不是對伺服器本身的評價,也不表示伺服器安全或目前可用:可連通性是帶日期的檢查結果,評分也不包含安裝量或使用者評價。

  • 近 30 天內可連通25 分

    請保持端點能夠回應 MCP 握手;SourceWeft 每 30 天探測一次。

  • 已宣告認證方式10 分

    請在註冊中心條目中宣告憑證請求標頭,或保持端點能夠回應探測,以便判定其認證方式。

  • 已列出工具15 分

    請讓伺服器在無憑證時回應 tools/list,或發布可在沙箱中無憑證啟動的軟體包。

  • 描述必需15 分
  • 圖示10 分
  • 文件15 分
  • 授權條款10 分

其他 Publish 整合