Usage instructions
Core usage
This Skill interacts with the locally installed Tencent IMA (ima.copilot) desktop application through Python scripts, and uses the Chrome DevTools Protocol (CDP) protocol to achieve automated control of the client. Users can trigger via natural language commandsima_searchTools to perform public web searches or precise searches against personal private knowledge bases. For the private knowledge base function, users need to obtain and configure it through packet capture in advance.knowledge_id, then used in the query@or@knowledgeprefix to activate exclusive knowledge retrieval mode. Script supportautocloseThe parameter controls whether to automatically close the application after searching. It remains on by default to optimize the continuous query experience.
Significant advantages
Security is the outstanding advantage of this Skill. All code execution is completed locally. The risk of command injection is completely eliminated through subprocess calls in the form of parameter lists, and there are no dangerous functions such as eval/exec. Network communication is strictly limited to the CDP debug port (127.0.0.1:8315) of the local loopback address, ensuring that user query data is not uploaded to third-party servers and only interacts with the official IMA client. The functional design is transparent and controllable. Private knowledge base access requires users to actively configure credentials. There is no silent data collection behavior, and config.json has been added with .gitignore to protect privacy. In addition, the 90-second timeout mechanism and complete error handling ensure execution stability and avoid resource occupation due to process suspension.
Potential Disadvantages and Limitations
Platform compatibility is limited, the current implementation only supports macOS systems and must rely on pre-installation of the official IMA desktop client. There are technical barriers to the configuration process of private knowledge bases. Users need to have certain network packet capture capabilities to obtain Knowledge IDs, which is not user-friendly enough for ordinary users. The function is limited by the locally applied CDP protocol. If the IMA client is not started or the port is occupied, the call will fail. In addition, the script implements functions by intercepting local network requests. Although it is safe, it is an unofficially supported automation solution. It may become invalid with the IMA client update and depends on the stability of a specific version.
Suitable target group
It is mainly intended for heavy users of Tencent IMA, especially professionals who have accumulated a large amount of personal knowledge assets in the IMA knowledge base. Ideal for developers and power users who care about data privacy and want to integrate AI search capabilities into native automated workflows (such as through Clawdbot or Claude). It is of high value to knowledge workers and researchers who need to frequently retrieve information in private documents, as well as technology enthusiasts who want to realize a closed loop of "local AI assistant + private knowledge base". It is not recommended for ordinary users who have no technical background, cannot complete the configuration packet capture steps, or do not understand the CDP protocol.
Risks of use
In terms of performance, each call needs to start or connect to the local IMA application, which has a 90-second timeout limit and is not suitable for real-time query scenarios with high frequency and low latency requirements. In terms of dependency risks, this solution relies on a specific version of the IMA client and CDP protocol. Official application updates may cause the script to become invalid, and continuous attention is required to maintain compatibility. Configuration risks include Knowledge ID as the access credential for the private knowledge base. Improper storage may lead to the leakage of knowledge base content. It is recommended to properly protect the config.json file to avoid credential leakage. In addition, error information is output to stderr. When deploying in a production environment, you need to pay attention to the log collection and monitoring configuration, as well as the occupation of local port 8315.