Google Workspace CLI (gws)
googleworkspace/cliA command-line tool for Drive, Gmail, Calendar, Sheets, Docs and Chat that builds its commands from Google's Discovery Service, plus about 40 thin agent skills.
With caveats
Check first Depth 2/5
Lark CLI (lark-cli) Claude skills
The official Lark/Feishu command-line tool with 200+ commands and about 29 agent skills for messaging, docs, Base, sheets, calendar, mail, tasks and more.
Our verdict Recommended
An official, actively maintained and unusually detailed pack for Lark and Feishu; the main thing to weigh is that the agent acts with your user permissions.
What Lark CLI (lark-cli) does
lark-cli is a Go command-line client for the Lark/Feishu Open Platform, and the repository ships the skills that teach an agent to drive it. Commands come in three layers: human- and agent-friendly shortcuts (names starting with +), API commands synced from the platform, and a raw API escape hatch. Covered areas include messenger, docs, drive, Base tables and dashboards, sheets, slides, calendar, mail, tasks, wiki, contacts, meetings and approvals.
The skills folder has one skill per area (lark-im, lark-doc, lark-base, lark-sheets, lark-mail, lark-calendar, lark-task and others), a lark-shared skill that handles app config, login, identity and scope rules for all the others, a skill-maker, and two workflow skills for meeting summaries and standup reports. They are detailed: about 9 KB per SKILL.md on average, backed by roughly 500 files in total, including per-command reference pages and Python helper scripts for sheets (profiling tables, checking charts) and slides (XML validation, icon lookup). Many skill texts, including lark-doc, are written in Chinese. The skills tell the agent to read only the reference for the scenario at hand, to treat email content as data rather than instructions, and to use dry-run and confirmation flags before destructive commands. Setup needs a Lark or Feishu app and an interactive OAuth login.
Good for
- Teams that live in Lark or Feishu and want an agent to read docs, send messages or manage Base data
- Automating calendar, task and meeting-summary workflows through the CLI
- Enterprises that want to embed the CLI in their own agent with central credentials
Not for
- Anyone not using Lark or Feishu
- Users who do not want an agent acting under their own Lark identity
What's inside depth 4/5
About 29 skill folders averaging 9 KB of instructions with hundreds of per-command reference pages and Python helper scripts for sheets and slides, though the skills ship no evals of their own.
- Skills
- 37
- Reference docs
- 428
- Scripts
- 17
- Tests / evals
- 3
- Templates
- 7
- Avg SKILL.md
- 8.6 KB
Counted from the repository file tree. How we score depth.
Dependencies and costs
- A Lark or Feishu account and a configured app (lark-cli config init)
- Node.js for npx install; Go 1.23+ and Python 3 only to build from source
- lark-whiteboard-cli via npx for whiteboard rendering
- Telemetry: On by default: requests to official Feishu/Lark HTTPS domains carry risk-control signals limited to OS type and device hardware model. Disable per workspace with
lark-cli config risk-control off. No other telemetry found.
Security review
Review before installingSkills for operating Lark/Feishu (docs, sheets, slides, mail, IM, calendar, tasks, wiki, apps) through the separate lark-cli, which then acts as you on your workspace once you authorize it. The skills carry explicit safety rules (for example, confirm before sending mail and treat email content as untrusted), but the CLI itself, its installer and its default-on risk-control signals are outside the files we could review.
Worth knowing
- Relies on a CLI or package we could not reviewAll skills drive the separate
lark-clibinary (Go, published as @larksuite/cli); the helper scripts shell out to it with fixed argument lists, and the CLI's own source was not among the files we had.skills/lark-sheets/scripts/lark_sheet_read_cli.py:59 - Makes network requestslark-cli acts under your Lark/Feishu user identity on the Open Platform (docs, drive, mail, IM, calendar, wiki, tasks, apps and more); after you authorize scopes the agent can read, send, publish and delete within them, with confirmation rules written into individual skills such as lark-mail.README.md:297
- Sends usage data by default (can be turned off)By default the CLI adds risk-control signals (OS type and device hardware model) to requests to official Feishu/Lark domains; turn off with
lark-cli config risk-control off.README.md:305 - Runs remote code at an unpinned versionInstall is
npx @larksuite/cli@latest installandnpx skills add larksuite/cli -y -g, unpinned; the CLI also reports updates via_noticeand tells the agent to suggestlark-cli update.README.md:75 - Installs packages or toolsThe installer fetches the lark-cli binary and writes the AI skills into the agent's skills folder; lark-whiteboard rendering uses lark-whiteboard-cli via npx.README.md:75
- Changes files outside your project
lark-cli config initandauth loginstore app credentials and user tokens in the OS keychain/CLI config on your machine.README.md:94
What we checked
The repository (3332 files) on Oct 29, 2026. Not a full review: 2807 of 3332 files were not downloaded (large repository; we reviewed the skill folders); we read 3 of 17 code files; it relies on code outside the repository: lark-cli Go binary / @larksuite/cli npm package and installer (source not in our files); lark-whiteboard-cli (npx); 2807 of 3332 repository files were not available to us (the Go source tree, workflows); the larger helper scripts (xml_lint.py, sxsd_validator.py, lark_chart_quality_check.py, deck-stage.js and others) were only searched for network and process use, not read in full.
NVIDIA SkillSpector static scan: 218 pattern hits (Prompt Injection 121, MCP Rug Pull 28, Privilege Escalation 19, Excessive Agency 16, Tool Misuse 10, Dangerous Code Execution 7). The 218 SkillSpector hits are mostly documentation: DELETE endpoint docs for wiki and drive commands, --no-validate flags, 'access token' notes, an anti-injection rule in lark-mail that quotes 'Ignore previous instructions' as something the agent must refuse, and os.execv in a test shim. deck-stage.js is a front-end slide component that fetches a local href; it does not call external hosts.
A review, not a certification. Skills change after we check them — read the files before installing.
License
MIT — commercial use allowed.
Can it run unattended?
Not really. The skills are interactive command guides with confirmation steps for risky writes; they include no long-running loop or resumable job state.
Skills in this repository
| Skill | Installs |
|---|---|
| lark-doc | 481K |
| lark-base | 479K |
| lark-shared | 478K |
| lark-im | 475K |
| lark-drive | 475K |
| lark-sheets | 475K |
| lark-wiki | 474K |
| lark-whiteboard | 473K |
| lark-task | 472K |
| lark-calendar | 472K |
| lark-mail | 470K |
| lark-minutes | 469K |
| lark-vc | 469K |
| lark-event | 469K |
| lark-contact | 469K |
| lark-workflow-meeting-summary | 468K |
| lark-workflow-standup-report | 467K |
| lark-openapi-explorer | 467K |
| lark-skill-maker | 467K |
| lark-approval | 443K |
| lark-slides | 423K |
| lark-attendance | 420K |
| lark-okr | 407K |
| lark-markdown | 388K |
| lark-vc-agent | 358K |
| lark-apps | 318K |
| lark-note | 245K |
Linked skills have their own review.
Alternatives to Lark CLI (lark-cli)
Obsidian Skills
kepano/obsidian-skillsSix skills by Obsidian's Steph Ango (kepano) for Obsidian Markdown, Bases, JSON Canvas, the Obsidian CLI, Defuddle and Knap.
Recommended
Check first Depth 2/5
Anthropic Agent Skills
anthropics/skillsAnthropic's reference skills repo: 20 skills including document tools (docx, pptx, xlsx, pdf), frontend-design, skill-creator and mcp-builder.
Recommended
Check first Depth 4/5
FAQ
What is lark-cli?
lark-cli is the official command-line tool for Lark and Feishu, maintained by the larksuite team. It offers 200+ commands across messaging, docs, Base, sheets, calendar, mail and tasks, plus agent skills that let AI coding agents run those commands.
How do I install lark-cli and its skills?
Run npx @larksuite/cli@latest install, then lark-cli config init and lark-cli auth login --recommend. If you build from source, also install the skills with npx skills add larksuite/cli -y -g, as the README describes.
Is lark-cli safe to use with an AI agent?
It stores credentials in the OS keychain and the README documents input-injection protections, but an authorized agent acts under your Lark identity and can read and send messages, so the README advises against adding the bot to group chats. The skills instruct the agent to treat email content as data and to confirm destructive actions.
lark-cli vs Google Workspace CLI: which should I use?
They serve different suites: lark-cli is for Lark and Feishu, and googleworkspace/cli is for Google Drive, Gmail and Calendar. Pick the one that matches the workplace tools you actually use.
Sources: GitHub repository (stars, license, file tree, last push Oct 29, 2026); install counts from skills.sh via Best Skills (CC BY 4.0). Review based on the README and SKILL.md files as published.