Skip to content

Latest commit

 

History

History
71 lines (45 loc) · 2.1 KB

File metadata and controls

71 lines (45 loc) · 2.1 KB

MrRSS Skills

English | 简体中文

MrRSS ships a Codex skill package for users who want an AI agent to inspect and operate their MrRSS data through the local REST API.

What Is Included

The release asset MrRSS-<version>-skills.zip contains:

  • mrrss-assistant/SKILL.md - agent workflow and safety rules.
  • mrrss-assistant/references/api.md - generated API reference from docs/SERVER_MODE/swagger.json.
  • mrrss-assistant/scripts/generate_api_reference.py - maintenance script for regenerating the API reference.
  • mrrss-assistant/agents/openai.yaml - Codex UI metadata.

Install

  1. Download MrRSS-<version>-skills.zip from the GitHub release page.
  2. Extract it.
  3. Copy the extracted mrrss-assistant folder into your Codex skills directory:

Windows:

Copy-Item -Recurse .\mrrss-assistant "$env:USERPROFILE\.codex\skills\"

macOS or Linux:

cp -R ./mrrss-assistant ~/.codex/skills/
  1. Restart Codex so it can discover the new skill.

Use

Start MrRSS in server mode or run the desktop app with its local API available.

Server mode:

docker run -p 1234:1234 ghcr.io/wcy-dt/mrrss:latest-amd64

Then ask Codex:

Use $mrrss-assistant to inspect my unread MrRSS articles and summarize the most important items.

The skill defaults to http://localhost:1234/api. If your instance uses another host or port, include it in the prompt.

Safety Model

The skill tells agents to:

  • prefer read-only API calls first;
  • ask before destructive operations, bulk status changes, cache clearing, or settings updates;
  • redact credentials and API keys from responses;
  • use MrRSS API endpoints instead of direct SQLite access unless the user explicitly asks for offline inspection.

Maintainers

When API routes change, regenerate the reference:

python skills/mrrss-assistant/scripts/generate_api_reference.py docs/SERVER_MODE/swagger.json skills/mrrss-assistant/references/api.md

CI checks that the generated reference stays in sync. The release workflow packages the skill as MrRSS-<version>-skills.zip and uploads it as a release asset.