learnprompt/carl-weread

carl-weread is a tool that plugs into WeChat Reading (the Tencent reading app called WeRead) and uses AI agents to change how you decide what to read. The official WeRead app already tells you how many minutes you have read, which books are on your shelf, and what your notes say. carl-weread tries to push one step further: given a problem you are stuck on right now, it picks one book on your shelf and one short section of that book to read today, then turns the reading into an action card. The core method has three pieces according to the README. First, it asks what kind of problem you are dealing with, for example whether you lack information or lack a frame to close a question, and chooses a reading goal that fits. Second, it cross-references several WeRead data sources together, including your bookshelf, your notebooks, chapter tables, reading progress, your own highlights, popular highlights, and the recommendation and similar-book endpoints. Third, instead of producing a weekly stats page, it asks which highlights actually turned into an article or a project decision and which books were only saved to ease anxiety. The project ships as a skill for AI agents using the Hermes skill format. A light install pulls in only the SKILL.md description so an agent can understand the method. A full install clones the repository, sets up a Python virtual environment, runs the tests, and lets you call scripts such as today_live.py, digest_apply.py, weekly_loop.py, and a unified carl_weread.py CLI with subcommands like recommend, today, after-read, and weekly. The WeRead API key is stored in a private file with mode 600 and is not committed to the repo. It supports four context modes: Obsidian vault, plain Markdown folder, chat only, and WeRead only. Output reading cards can be written back into your notes folder for later weekly review.