Use the current feishu_* tools as the source of truth for Feishu Project (Meego).
Load only the reference matching the requested branch.
1. Resolve the locator
Prefer a user-provided Meego URL. When a tool accepts url, pass the URL directly so it
extracts project_key, work-item type, item ID, or view ID. Otherwise resolve the space with
feishu_search_project_info.
If no space is present in the request or conversation, inspect only matching key lines in these
project-local files, in order: .agents/constant.md, .agents/constants.md, local.properties,
.env, .env.local. Recognize feishu-project-key, feishu_project_key,
FEISHU_PROJECT_KEY, or project_key. Never expose unrelated values. Ask for the space only
when no candidate is available.
Resolve uncertain or custom work-item types with feishu_list_workitem_types; do not guess a
key from its display name.
2. Route by branch
| Branch | Load |
|---|---|
| Space, types, fields, roles, relations | references/space-metadata.md |
| Work-item query/create/update/state | references/work-items.md |
| Views, charts, fixed and panorama views | references/views-charts.md |
| Nodes, subtasks, node transitions | references/nodes-subtasks.md |
| Teams, users, schedules/workload | references/teams-users.md |
| Comments, operation history, man-hours | references/activity.md |
| Resource libraries and deliverables | references/resource-work-items.md |
| WBS plan drafts and published rows | references/wbs.md |
| Upload/download and rich-text files | references/attachments.md |
| Complex MQL searches | references/mql-guide.md |
references/tool-reference.md is the compact inventory when the correct branch is unclear.
3. Apply schema-first preparation
- Create work item: call both
feishu_list_workitem_field_configandfeishu_list_workitem_role_config; obtain the mandatory template ID from thetemplatefield options beforefeishu_create_workitem. - Update field/role: resolve the exact field or role config first. If fuzzy lookup returns multiple candidates, ask the user to choose instead of guessing.
- Create resource instance: call
feishu_get_resource_work_item_type_conffirst. - Node fields: use
feishu_list_node_field_config; node schedules and owners follow the preservation rules inreferences/nodes-subtasks.md. - MQL: identify the target space and target work-item type. Query field config for field semantics, role config for roles, and use relation names exactly as the user supplied.
- WBS: distinguish draft rows from published instance rows. Mutations target a draft and must be followed through their operation status.
Treat each live tool schema and returned error as authoritative. On a schema/query error, adjust only the failing field, relation, parameter, or clause. An empty successful query is a valid result.
4. Pagination and completion
Pagination is endpoint-specific:
feishu_list_todo: 50 rows/page; fetch pages 1, 2, … only when the user requests all results, stopping when a page has fewer than 50 rows or no data.feishu_search_by_mql: reusesession_id; pass an emptymqlon later pages and paginate by returnedgroup_id.- WBS row APIs: continue while
has_moreis true. - Team, field/role, view, comment, and activity APIs use their returned page token/count semantics.
Do not apply one generic page-size rule to every endpoint.
5. Mutations and results
Execute the requested mutation once prerequisites are known. Full WBS publication is the one
operation requiring the exact confirmation in references/wbs.md; partial publication does not.
Present compact, relevant fields rather than a fixed table shape. Include item URLs when returned. For failures, report the tool error and the unresolved input; do not claim success from an accepted asynchronous operation until its progress endpoint confirms completion.