Use lark-cli to track research phase status and mirror research files when the user wants visibility inside Lark Docs and Base.
lark-cli config init --new is refused inside a Hermes context because it would create a parallel app and shadow the existing Hermes binding.lark-cli config bind.bot-only = safer default, no impersonation, no personal calendar/mail/drive access.user-default = user identity allowed, needed for personal-resource access.Avoid duplicating the research content inside Lark Base. Instead, maintain three clean layers:
- Source of truth: local RES_YYYYMMDD_Name.md file. Always up to date, detailed, and formatted for AI reading.
- Mirror: a Lark Doc named after the topic (e.g. VietnamTourismData), updated after each phase to reflect the current state.
- Tracker: records inside the Lark Base Research Tasks table mapping to research phases, containing status, record link, and follow-up notes.
For CLI scopes that require browser approval (e.g., doc creation, modification), use the split-flow:
1. Trigger fresh login: lark-cli auth login --scope "docx:document:create" --no-wait --json
2. Extract URL and return control to the user.
3. Upon user's "done", resume: lark-cli auth login --device-code <device_code>
4. Note that scopes are cumulative (incremental). E.g., docx:document:create and docx:document:write_only must both be granted to support doc create and overwrite workflows.
lark-cli parameters such as --content @file reject absolute paths (e.g., /tmp/...). Always use relative paths starting from the current working directory.lark-cli can fail if the markdown contains shell-interpreted characters like & (confused with backgrounding). Prefer writing a temp file via write_file and reading it with @relative_path.