← Back to Skills Marketplace
278
Downloads
0
Stars
1
Active Installs
2
Versions
Install in OpenClaw
/install model-switch-notify
Description
模型切换通知。当agent使用的模型发生变化时，第一时间通知当前会话用户。支持心跳检测机制，消息中断时下次会话自动通知。使用 SQLite 存储。
Usage Guidance
Do not install or enable this skill until you confirm and fix the data-path discrepancy. Specifically: (1) Inspect scripts/check_model.py and change DATA_DIR to a path that uses the running user's home directory (e.g. Path.home()/'.openclaw/data' or Path('~/.openclaw/data').expanduser()). (2) Verify the script's behavior in a sandboxed environment to confirm it creates the database where you expect and only accesses that DB. (3) Ensure permissions on the data directory are appropriate so other users/processes cannot read or tamper with the DB. (4) After patching, run a few test invocations to confirm the JSON output and pending-notify behavior match the docs. If you cannot modify the script yourself, treat the hardcoded path as a blocker and avoid installing.
Capability Analysis
Type: OpenClaw Skill
Name: model-switch-notify
Version: 1.0.1
The skill is designed to track and notify users of AI model changes within an OpenClaw session. It uses a local SQLite database (`model-switch.db`) to persist model states and pending notifications. The Python script `check_model.py` implements standard argument parsing and safe, parameterized SQL queries. While the script contains a hardcoded absolute path (`/home/wljmmx/`) which limits portability and suggests a development oversight, there is no evidence of data exfiltration, malicious execution, or prompt injection. The behavior is entirely consistent with the stated purpose.
Capability Assessment
Purpose & Capability
Functionality (detect model changes, heartbeat, pending notifications) matches the skill description and the script uses a local SQLite DB which is reasonable for this purpose. However the code hardcodes DATA_DIR = /home/wljmmx/.openclaw/data while the docs and integration examples use ~/.openclaw/data — this mismatch is unusual and not justified by the documentation.
Instruction Scope
SKILL.md instructs the agent to invoke the included Python script as a subprocess and pass agent/model/session/channel — that scope is appropriate. But the instructions expect storage under the invoking user's home (~/.openclaw/data) while the script will create and use /home/wljmmx/.openclaw/data. That means the script will read/write a different filesystem location than the documentation claims, which is unexpected behavior and could lead to silent failures or files written to another user's path.
Install Mechanism
No install spec is provided (instruction-only with an included script). This is lower risk than fetching arbitrary remote code; the code is bundled locally so nothing is downloaded at install time.
Credentials
The skill requires no environment variables or credentials, which is appropriate. The remaining concern is the hardcoded absolute data path in the script — it implicitly requires filesystem write access at /home/wljmmx/.openclaw/data (or will create that path), which is disproportionate and inconsistent with the documented ~/.openclaw/data location.
Persistence & Privilege
The skill does persist state to a local SQLite database and is not marked always:true. It does not request elevated privileges or modify other skills. Persistence is limited to its own data file (though the exact path is the aforementioned concern).
How to Use
- Make sure OpenClaw is installed (local or Docker)
- Run the install command in chat:
/install model-switch-notify - After installation, invoke the skill by name or use
/model-switch-notify - Provide required inputs per the skill's parameter spec and get structured output
Version History
v1.0.1
**Major update: Switch to SQLite storage, add heartbeat detection, and interruption handling.**
- Replaced JSON file storage with SQLite for improved data management.
- Introduced heartbeat detection: checks model status and pending notifications on every reply.
- Added interruption handling: if notification fails, it's stored and sent at the next available opportunity.
- Expanded command-line tools for state check, heartbeat update, interrupt set, state fetch/list/reset.
- Updated integration and usage instructions for improved reliability and cross-session continuity.
v1.0.0
model-switch-notify v1.0.0
- Notifies users when the agent's model changes, at the first reply, after a model switch, or upon user request.
- Stores and checks previous model information to prevent redundant notifications.
- Provides customizable notification templates and format.
- Allows configuration of notification preference per agent via USER.md.
- Includes scripts and integration guide for easy adoption by all agents.
Metadata
Frequently Asked Questions
What is Model Switch Notify?
模型切换通知。当agent使用的模型发生变化时，第一时间通知当前会话用户。支持心跳检测机制，消息中断时下次会话自动通知。使用 SQLite 存储。 It is an AI Agent Skill for Claude Code / OpenClaw, with 278 downloads so far.
How do I install Model Switch Notify?
Run "/install model-switch-notify" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.
Is Model Switch Notify free?
Yes, Model Switch Notify is completely free, licensed under MIT-0. You can download, install and use it at no cost.
Which platforms does Model Switch Notify support?
Model Switch Notify is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).
Who created Model Switch Notify?
It is built and maintained by wljmmx (@wljmmx); the current version is v1.0.1.
More Skills