DingTalk CLI Skill Index

The dws command is provided by the Super Magic runtime. Its current default installation contains separate dingtalk-* skills under ~/.agents/skills/. Use the task-specific skill below; do not require the legacy all-in-one dws skill.

Choose a Skill

For a clear product request, load only the matching skill with read_skills and follow its reference and shared-rule requirements. For authentication, account selection, an ambiguous request, or a cross-product workflow, start with dingtalk-shared.

User intent	Load skill
Login, authentication status, accounts, profiles, or cross-product routing	`dingtalk-shared`
Read, search, send, reply to, forward, or recall messages; manage chats	`dingtalk-chat`
Find people by name, department, responsibility, or reporting lines	`dingtalk-aisearch`
Look up users, departments, and directory relationships	`dingtalk-contact`
Read, create, or edit DingTalk documents and document content	`dingtalk-doc`
List, upload, download, or organize cloud drive files	`dingtalk-drive`
Operate AI tables, fields, records, views, charts, or dashboards	`dingtalk-aitable`
Manage calendar events, attendees, meeting rooms, or availability	`dingtalk-calendar`
Read or send mail	`dingtalk-mail`
Query meeting minutes, transcripts, summaries, or action items	`dingtalk-minutes`
Manage todos	`dingtalk-todo`
Manage wiki spaces, nodes, or members	`dingtalk-wiki`
Listen for future messages, approval changes, or other real-time events	`dingtalk-event`
Create or update AI cards	`dingtalk-aicard`
Manage digital employees or their local agent connections	`dingtalk-tag`
Approvals, attendance, DING, reports, spreadsheets, whiteboards, contracts, goals, app development, OpenAPI, or other long-tail products	`dingtalk-misc`

read_skills(skill_names=["dingtalk-shared"])
read_skills(skill_names=["dingtalk-chat"])
read_skills(skill_names=["dingtalk-misc"])

Use the loaded skill's returned <skill_dir> to resolve reference files. Do not guess a per-product skill name from the CLI command: several products share dingtalk-misc.

Authentication in the Super Magic Sandbox

For a status-only request, use the local snapshot without refreshing credentials:

dws auth status --readonly --format json

A snapshot is not an online token-validity check. Before a business operation, follow the selected skill's authentication requirements. If login is needed, use the device flow because the user's browser cannot reach the sandbox's loopback listener:

dws auth login --device --no-browser

Start login with shell_exec. If it returns a background task ID, return the authorization link and required user code to the user before waiting. After the user confirms authorization, use shell_await to resume that same task. Do not start duplicate login attempts. --no-browser alone only suppresses browser launch; it does not replace --device. Verify the result with:

dws auth status --format json

Do not initiate login merely to answer whether the user is already logged in. Follow the shared account-selection rules; do not pick an arbitrary account. Never expose access tokens, client secrets, or credential files.

Command Discovery and Missing Skills

Inspect the installed CLI before choosing parameters:

dws --help
dws <product> --help
dws <product> <command> --help
dws schema --cli-path "<command path>" --compact --format json

Use only commands supported by the installed version. A CLI command being available does not prove its skill is installed. If a selected skill is missing, inspect the installed skill directories. Use legacy dws guidance only when that mono skill is actually installed. Otherwise report the missing runtime skill; do not install, upgrade, or change skill mode inside the sandbox.

Follow the selected skill's confirmation and retry rules. A zero process exit code alone does not prove the requested business operation completed.

dingtalk-cli

DingTalk CLI Skill Index

Choose a Skill

Authentication in the Super Magic Sandbox

Command Discovery and Missing Skills

Is this your skill?