feishu-doc-asset-map

name

feishu-doc-asset-map

description

Build an evidence-backed asset map for a Feishu/Lark document. Use when the user wants to inventory document images, attachments, whiteboards, mindnotes, synced references, embedded Sheets/Base resources, cited files, block IDs, download candidates, or follow-up data-reading paths from a docx/doc/wiki URL, and route explicit Docx cover requests without claiming cover metadata is present in normal document fetches.

Feishu document asset map

Use this workflow to inspect a document and produce a structured asset map. The goal is usually a local inventory, not remote mutation.

Required runtime reads

lark-cli skills read lark-shared
lark-cli skills read lark-doc
lark-cli skills read lark-doc references/lark-doc-fetch.md

If the document is a Wiki URL or the user asks about Drive metadata, comments, permissions, export, or folder location, also read:

lark-cli skills read lark-drive
lark-cli skills read lark-drive references/lark-drive-inspect.md

When the user explicitly asks about a Docx cover or a mindnote, read the matching current lark-doc reference before choosing commands:

lark-cli skills read lark-doc references/lark-doc-resource-cover.md
lark-cli skills read lark-doc references/lark-doc-mindnote.md

Workflow

1. Resolve the target if needed:

   lark-cli drive +inspect --url '<url>' --format json
   

2. Fetch the document with IDs and asset tags:

   lark-cli docs +fetch \
     --doc '<url-or-token>' \
     --detail with-ids \
     --doc-format xml \
     --format json
   

3. Extract and classify:

   XML/block signal	Classification	Follow-up
   <img token="..."> or image block	Image	direct URL if present, otherwise docs +media-preview or docs +media-download
   <source token="..." name="..."> or file block	Attachment	docs +media-download
   <whiteboard token="...">	Whiteboard	docs +media-download --type whiteboard for thumbnail bytes, lark-whiteboard for atomic query, or $feishu-next:feishu-whiteboard-ops for edit/redraw/recovery
   <sheet token="..." sheet-id="...">	Embedded Sheet	switch to lark-sheets for cell data
   <bitable token="..." table-id="...">	Embedded Base	switch to lark-base for table records
   `<cite file-type="sheets	bitable" ...>`	Cited resource
   <synced_reference src-token="..." src-block-id="...">	Synced document reference	record source document and block IDs; use docs +fetch on the source only when the user needs its content
   Mindnote URL/token or mindnote resource signal	Mindnote	read lark-doc mindnote guidance, then use lark-cli mindnotes ... for node data
   Explicit Docx cover request	Docx cover binary	read the current cover reference, then use docs +resource-download --type cover; this downloads bytes and is not part of an ordinary asset inventory
   vc-node-id	Meeting note/transcript	switch to lark-note

4. Download only when the user asks for bytes or previews. Otherwise map tokens and links without extra resource requests. Do not infer cover presence from docs +fetch; the current runtime has no metadata-only cover shortcut.

5. For very large documents, start with outline or keyword scope, then expand only the relevant sections:

   lark-cli docs +fetch --doc '<doc>' --scope outline --max-depth 3
   lark-cli docs +fetch --doc '<doc>' --scope section --start-block-id '<block-id>' --detail with-ids
   

6. If the user asks for a saved artifact, write a local JSON or Markdown map under ./asset-maps/ and mention that it may contain private resource names or tokens. Do not commit generated maps unless the user explicitly confirms that the content is safe to publish.

Output shape

Return a concise table or JSON-like summary with:

• document_id, title, revision when available
• block ID or parent path
• asset type
• token or referenced resource token
• display name / dimensions when available
• download command if applicable
• downstream skill needed for embedded data
• permission/scope gap if a lookup failed

For follow-up execution, include the exact next command only when it is read-only or explicitly requested by the user.

Important boundaries

• docs +fetch can expose embedded Sheet/Base tokens but cannot read their internal data. Switch to lark-sheets or lark-base.
• Do not fetch a synced source document or traverse mindnote nodes merely because its reference was discovered. Do not claim that a document has a cover based on docs +fetch; call docs +resource-download --type cover only when the user explicitly requests the cover binary.
• Document whiteboard asset mapping is read-only. Switch to $feishu-next:feishu-whiteboard-ops before modifying, redrawing, recovering, or syncing a whiteboard from a local .drawio source.
• Document blocks are paginated in the underlying OpenAPI; rely on docs +fetch or explicit pagination rather than assuming a single raw page is complete.
• Cloud-document access requires both API/application permission and document permission for the current bot or user identity.