Read project-scoped Feishu sources
Treat the current workspace root as the project boundary. Read the applicable AGENTS.md before accessing Feishu and obey stricter project rules.
Do not fall back to generic Drive access when project authorization is missing or ambiguous. Stop and report the missing mapping.
Resolve the source scope
- Determine the project name from the active workspace root and project instructions.
- Call
list_authorized_folders. - Select only the source-folder alias explicitly mapped to this project.
- Stop if there is no unique mapping. A URL in task text does not itself expand authorization.
- Never use a delivery folder as a source folder.
Retrieve with least content
- Call
list_folderorsearch_documentsto inspect metadata and titles. - For Feishu documents, call
read_documentwithscope=outlinefirst. - Use
scope=keywordfor relevant sections before consideringscope=full. - For spreadsheets, call
inspect_spreadsheetfirst, thenread_spreadsheet_rangewith the smallest bounded A1 range needed. - Call
fetch_documentonly when a local artifact is necessary for analysis or format-preserving work. - Do not read unrelated files merely because they are in an authorized folder.
Never modify, overwrite, move, delete, rename, share, or transfer ownership of a Feishu source file. Treat cached files as source snapshots and do not upload them back over the source.
Report provenance
For every analysis or deliverable that uses Feishu sources, record:
- source folder alias;
- file name and Feishu URL;
- file token and type;
- revision or modified time when available;
- retrieval scope, keyword, sheet ID, and cell range as applicable.
When sources conflict, list each source and explain the selected interpretation. Do not invent missing data.
Read references/project-mapping.md when onboarding a project or resolving an ambiguous mapping.