work-review-timemachine for feishu

Summarize a user's recent Feishu work into a concise weekly review or "time machine" recap by reading Feishu docs, meeting notes, and related materials, then...

Security Scan

The instructions clearly ask the agent to read and create Feishu docs (accessing user workspace data) but the skill declares no credentials, config paths, or install steps to do that, creating an incoherence you should resolve before installing.

Purpose & Capability

The skill's stated purpose is to read a user's Feishu docs/meeting notes and create/update Feishu docs. Yet the package declares no credentials, no config paths, and no required binaries. Reading and writing a user's Feishu workspace normally requires an API token/connector or explicit auth flow, so the declared requirements do not match the functionality described.

Instruction Scope

SKILL.md instructs the agent to search the user's recent docs/wiki nodes, prioritize docs owned by the user, read transcripts, and create/update Feishu docs. Those runtime actions imply access to user data and writable workspace privileges. There is no explicit limit on which docs to read beyond heuristics, and no explicit mention of where collected data may be sent (it implies internal Feishu operations). The instructions also reference handling permission cards — which suggests an interactive OAuth/consent flow not documented in the skill manifest.

Install Mechanism

This is instruction-only with no install spec or downloadable code, so nothing new is written to disk. That lowers installation risk. There are no suspicious install URLs or packages.

Credentials

No environment variables, credential fields, or config paths are declared, yet the skill requires reading and writing Feishu workspace content. Either the skill expects the platform to supply Feishu connector credentials (not documented), or required credentials were omitted. Absence of declared secrets for a capability that needs them is disproportionate and should be clarified.

Persistence & Privilege

The skill does not request always:true, does not install binaries, and does not claim persistent agent-level modifications. Autonomous invocation is allowed (platform default) but not combined with other elevated privileges in the manifest.

What to consider before installing

This skill's instructions require reading and creating Feishu documents, which normally needs explicit authorization (API token or platform connector). Before installing, ask the publisher or platform: (1) how will the agent obtain Feishu access — is there a built-in Feishu connector or must you supply credentials/consent? (2) what exact OAuth scopes or API tokens will be used (read-only vs. write), and can you limit them? (3) where does any gathered data go — is it stored externally or only used in-session? If the platform provides a managed Feishu integration, verify its scope and consent flow; if not, treat the lack of declared credentials as a red flag and avoid granting broad workspace access until you confirm the auth mechanism and minimum required scopes. Consider testing with an account that has limited permissions or with non-sensitive docs first.

Feishu Work Review Timemachine

Overview

Create a compact, high-signal review of a user's recent work from Feishu materials. Focus on what the user actually drove or owned, not everything discussed in meetings.

Workflow

1. Define scope first

Clarify or infer:

• time range: last week / last month / specific dates
• output form: concise recap / weekly breakdown / monthly summary / boss-facing version
• destination: chat reply only, or create/update a Feishu doc

If the user already gave a clear scope, do not ask again.

2. Gather source materials from Feishu

Prefer this order:

1. Search the user's recent docs/wiki nodes
2. Prioritize docs created by or owned by the user
3. Read key docs in full only when needed
4. Prefer summaries / structured notes first, then raw transcript docs for validation

Useful source types:

• meeting notes and transcripts
• work-in-progress docs
• project plans
• technical方案 / design docs
• quality / metrics docs
• wiki pages tied to ongoing initiatives

Do not try to read everything. Read enough to identify repeated themes and evidence of ownership.

3. Separate real work from meeting noise

Do not treat every meeting mention as the user's work.

Use these signals to infer stronger ownership:

• the user is the doc owner / creator / primary editor
• the user is the main speaker framing goals, tradeoffs, or next steps
• action items are explicitly assigned to the user
• the user is pushing the plan, organizing follow-up, or coordinating POCs
• the same initiative appears across multiple docs with the user in a central role

Treat these as weaker signals:

• the user merely attended the meeting
• another person's project status appears in notes
• broad team discussion without assigned ownership
• AI-generated meeting summary claims not supported elsewhere

When uncertain, downshift wording:

• use “参与推动 / 协同推进 / 讨论并判断”
• avoid overstating as “主导完成”

4. Extract only the core themes

Compress the material into a few workstreams, usually 3-5:

• platform / product / project track
• AI / efficiency / technical exploration track
• metrics / governance / methodology track
• cross-team coordination track

For each workstream, capture only:

• what changed
• what the user specifically drove
• why it mattered

Delete or skip:

• optional commentary
• long caveats
• repeated explanations
• every subtask from meeting minutes
• alternate interpretations unless the user asked for analysis

5. Write in layers

Default output order:

1. one-paragraph core conclusion
2. 3-5 core workstreams
3. weekly breakdown if requested
4. one-sentence summary

If the user says “keep only the core”, reduce to:

• core conclusion
• core workstreams
• optional one-line weekly view

Output pattern

Compact monthly review

Use this structure:

# 时光机回顾｜过去一月主要工作总结

## 核心结论
[1 short paragraph]

## 核心主线
### 1. [主线]
- [核心动作]
- [核心动作]

### 2. [主线]
- [核心动作]

## 按周回顾
### 第 1 周：[关键词]
[1-3 lines]

### 第 2 周：[关键词]
[1-3 lines]

## 一句话总结
[1 sentence]

Boss-facing version

Make it shorter and more outcome-oriented:

• progress
• scope / influence
• cross-team traction
• next-step signal

Avoid process-heavy wording.

Feishu doc creation/update

When the user asks to create or write a Feishu doc:

• create the doc once a solid draft exists
• if permissions block the action, tell the user to complete the authorization card and then resume
• once created, prefer overwriting the draft if the user asks to simplify aggressively
• when refining, keep the same doc unless the user asks for a new one

Style rules

• Write in concise Chinese unless the user asks otherwise
• Prefer short sections over long prose
• Remove “可选项” and hedging text unless uncertainty is material
• Use strong verbs for high-confidence ownership, softer verbs for medium-confidence involvement
• Default to fewer bullets, not more
• Keep the final document skimmable in under 1-2 minutes unless the user asked for depth

Anti-patterns

Do not:

• dump a list of all meetings
• copy AI meeting summaries verbatim
• equate attendance with ownership
• preserve every nuance when the user wants a crisp recap
• keep “next-step suggestions” if the user asked for a finished summary

Final check

Before delivering, verify:

• Is this mostly about the user's actual work, not everyone else's?
• Did I collapse repeated themes into a few clear workstreams?
• If asked to simplify, did I remove optional sections rather than just shorten sentences?
• Would the user immediately recognize this as accurate and useful?