Versioned domain experts and App Bundles for DashScope AgenticX 1.0. This repository contains application profiles and Skills, not a copy of the AgenticX Runtime.
All bundles live under apps/<name>/. There are no legacy top-level Expert
directories or legacy profile layouts.
| App | Purpose |
|---|---|
basic-chat |
Minimal conversation profile and reusable teaching Skills |
dashscope-sdk-expert |
DashScope SDK usage, diagnosis, and API migration |
bailian-expert |
Bailian application design, configuration, and diagnosis |
Each bundle contains app.toml, a self-contained .agenticx/ profile,
evaluation metadata, and acceptance configuration. The SDK Expert includes
its complete domain Skills; it does not depend on a sibling Expert directory.
Each profile includes .agenticx/expert.toml with its Expert name. AgenticX
uses this record to display the selected Expert in its welcome panel. The
canonical role-template commands are /expert and agenticx expert.
Requires Python 3.10+ and AgenticX Runtime 1.0.0.
python -m pip install "dashscope-agenticx>=1.0.0,<2"
git clone https://github.com/dashscope/dashscope-agenticx-experts.git
cd dashscope-agenticx-experts
agenticx app install ./apps/dashscope-sdk-expert
agenticx app info dashscope-sdk-expert
agenticx app verify dashscope-sdk-expertInspect the App's manifest and requested permissions before installation. A checksum verifies content identity, not publisher authenticity. An App's publisher field alone does not confer official trust.
For a teaching workspace, the Runtime can download a profile from apps/
into the current workspace's .agenticx/ directory:
export DASHSCOPE_API_KEY="<your-api-key>"
agenticx expert download basic-chat
agenticxWorkspace scaffolding is not a substitute for versioned App installation or permission review. Keep credentials in environment variables or the OS keyring; never put real secrets in committed profiles.
With AgenticX Runtime 1.0.0 installed:
python -m pip install pytest
python -m pytest -q tests/test_apps.pyThe catalog tests check manifests, runtime requirements, evaluation and acceptance files, self-contained profiles, and the absence of legacy top-level Expert directories. They do not call model APIs.