Documents

Lark Work Report

Try it

Preflight-check Lark channel and send completion reports

What it does

Use this skill when the user asks to: report completion/progress to Feishu/Lark establish or repair a stable "work group" reporting channel before task execution

The skill document

Lark Work Report

Use this skill when the user asks to:

  • report completion/progress to Feishu/Lark
  • establish or repair a stable "work group" reporting channel before task execution

Required MCP interfaces

  • Find group: im_v1_chat_list (useUAT=true)
  • Send report: im_v1_message_create (useUAT=true, receive_id_type=chat_id)
  • Create work group if needed: im_v1_chat_create (useUAT=true)
  • Resolve user identity when needed: contact_v3_user_batchGetId (useUAT=true)

Default work-group bootstrap

Target group name: Codex工作群

  1. Run im_v1_chat_list and search exact group name match (name == "Codex工作群").
  2. If exactly one match exists, store its chat_id and send a short connectivity test message.
  3. If no exact match:
    • list likely candidates from im_v1_chat_list (names containing Codex or 工作群)
    • ask user to confirm whether one candidate should be used
  4. If still unresolved:
    • ask user for identity details (name/mobile/email/open_id)
    • ask if a new Codex工作群 should be created
    • create group via im_v1_chat_create, send a test message, and wait for user confirmation
  5. Send one test message to the selected/new group and confirm with the CLI user that the channel is correct.
  6. Only after reporting channel is confirmed, proceed with the main task workflow.

Reporting workflow

After finishing a user task:

  1. Send a concise report message to the resolved work-group chat_id.
  2. Report should include:
    • task summary
    • key outputs/artifacts
    • blockers/risks (if any)
    • next-step suggestion (if any)
  3. Keep message text compact. If content is long, split into multiple sequential messages.

Explicit non-goals

  • This skill does not monitor Feishu for follow-up instructions.
  • This skill does not run polling loops for new tasks.
  • Task intake/clarification happens in CLI conversation unless the user explicitly requests another workflow.

Safety and behavior constraints

  • Never assume group identity from memory; always verify via API in the current session.
  • If sender identity or group identity is ambiguous, confirm with the user before sending task reports.
  • Do not silently switch to another group without explicit user confirmation.

Related skills

Find why your productivity system keeps failing, then apply the smallest fix — capacity math, bottleneck routing, durable local notes.

by Iván855 installs69 stars

Generate and edit Draw.io, Mermaid, and Excalidraw diagrams from natural language using a structured JSON spec.

by nssa.io1.0k installs47 stars

Join a video meeting as an AI bot with voice, avatar, and screenshare across four operating modes.

by johnpatternai22 installs8 stars

Stores durable facts in a categorized, plain-markdown vault on disk, alongside your agent's built-in memory.

by Iván559 installs18 stars

Fetch raw ad creative, app, ranking, and revenue data from AdMapix as structured JSON.

by fly0pants4.3k installs296 stars

More from xukp20

Browse all skills

Model-led arXiv collection: author queries, judge relevance per result, then merge and dedupe into a final paper set.

by xukp2098 installs

Download ArXiv source or PDF artifacts per paper, read the paper, and write a structured summary.md in your chosen language.

by xukp20119 installs1 stars

Orchestrates end-to-end arXiv paper retrieval, processing, and batch reporting with language control and parallel or serial paper handling modes.

by xukp2057 installs

Generates a batch arXiv report by combining model-written templates with brief conclusions and abs URLs extracted from per-paper summary.md files.

by xukp2055 installs