AGENT SKILL

amoorion

Use amoorion to publish, read, reply, delete and search on chatting and communication platforms (Bluesky, Mastodon, X, Tumblr, Blogger, Facebook, Discord, Telegram, Matrix, ntfy, e-mail) from one CLI, Python object, HTTP API or MCP server; secrets come from a vault. Use when asked to post to several platforms, read what came in, answer it, automate reactions to messages, or check which platforms are set up.

License 许可
MIT
所属仓库
Sarverott/amoorion
最近更新
2026-11-05
Stars
★ 2
Forks
0

01 前言 Front Matter

name
amoorion
description
Use amoorion to publish, read, reply, delete and search on chatting and communication platforms (Bluesky, Mastodon, X, Tumblr, Blogger, Facebook, Discord, Telegram, Matrix, ntfy, e-mail) from one CLI, Python object, HTTP API or MCP server; secrets come from a vault. Use when asked to post to several platforms, read what came in, answer it, automate reactions to messages, or check which platforms are set up.

02 资源文件 100

03 SKILL.md 正文

amoorion

One tool for many platforms. Every operation is a method; the CLI, the Python object, the HTTP API and the MCP server all expose the same methods.

Before anything

  1. amoorion plugins: the platforms installed and what each can do. Each platform is its own package: pip install amoorion-bluesky.
  2. amoorion accounts: which platforms have their secrets in the vault. A platform without them refuses to work; never ask the user for a token in the chat, it belongs in their vault.
  3. Try with the platform dryrun first: it sends nothing.

Methods

Read-only, safe to run:

amoorion --json plugins                               # what is installed
amoorion --json accounts                              # which secrets are set (values never shown)
amoorion --json read --platform bluesky --limit 5     # latest messages; also kept in the inbox and dataset
amoorion --json inbox                                 # everything read or received so far
amoorion --json search --text "invoice"               # find in the inbox
amoorion --json snapshot --platform bluesky           # a profile's posts into the dataset
amoorion --json wiki --title "Cupid"                  # a Wikipedia summary
amoorion --json simulate --topic "…" --personas …     # personas react, locally; needs a language model
amoorion --json skill                                 # this text, and where the file is

They change something on a platform or reach people, so confirm first:

amoorion publish --text "hello" --to dryrun bluesky mastodon
amoorion reply --platform bluesky --to "<post id>" --text "…"
amoorion delete --platform bluesky --id "<post id>"            # a copy stays in the dataset
amoorion notify --text "…"                                     # to the platforms of AMOORION_NOTIFY
amoorion workflow --steps …                                    # methods in order, stops at the first failure
amoorion workflow --rules …                                    # rules on the inbox, once per message

amoorion compose --prompt "…" asks a language model (from the vault) to write something; it posts nothing. amoorion <method> -h tells the parameters of any method, and --json before the method gives machine-readable output.

From Python

from amoorion import Amoorion, Plugins

amo = Amoorion().use(Plugins["dryrun"])    # a sandbox: nothing of the real setup
amo.publish(text="hello")                  # every method by its name

real = Amoorion(".env").use(Plugins["bluesky"])    # the user's real vault, inbox and dataset

Amoorion.init() puts the whole process in a sandbox: also the CLI, Inbox() and Plugins, for what doesn't go through an object.

As servers

amoorion mcp runs the MCP server (stdio) with the methods as tools; the ones that change something on a platform are not marked read-only. amoorion api serves them over HTTP, amoorion daemon runs the long-running services.

Rules

  • Publishing, replying, deleting and notifying are visible to real people: get the user's confirmation of the text and the platforms before running them.
  • Secrets live only in the vault the user's .env links to; never print, log or commit them.
  • Docs: https://amoorion.readthedocs.io (Markdown of any page: ask with Accept: text/markdown, or read https://amoorion.readthedocs.io/llms.txt).

04 同仓库更多 Skills