Developer Open-Sources Sensitive Word Extraction Tool for WeChat Keyboard

On 2026-11-06 at 08:26, a developer posted on V2EX that they had built and open-sourced, with the help of AI, a tool for extracting the built-in sensitive word list from WeChat Input, the PC input method made by Tencent. According to the post, WeChat Input ships with an internal dictionary of sensitive terms baked into the installation package rather than fetched over the network: when an input matches a term in that dictionary, it cannot form compounds or associations, is not learned as a user memory, and never shows up in the candidate list. The poster demonstrated this behavior with an example, which is not reproduced here. The dictionary currently contains a little over 10,000 entries, the developer said. The tool reads the file .\BuildIn\cell_dict.ssrc.zh from the WeChat Input installation directory and writes the extracted entries to a text file. It was tested on the latest Windows version of WeChat Input and is run from the command line as python export_sensitive_single.py -f <cell_dict.ssrc.zh> -o <out.txt>, so users need only point it at the dictionary file and choose an output path. The download is hosted on pastebin at https://pastebin.com/emEXbrCh (access code x7E56zQ4uZ), with the tool aimed at technically inclined users interested in how input method dictionaries work. The developer also noted a second finding: WeChat Input carries an additional, less sensitive word list identified internally by kind=440. Entries in this secondary list can still form compounds, the poster wrote, but they are classified separately within the input method's implementation, and whether they receive any special handling, such as being logged, was left unexamined. As of the only available post, no comment or response from Tencent regarding the existence of the dictionary or the extraction tool has appeared.