Kingsoft Docs CLI

Problem It Addresses

When an agent operates Kingsoft Docs directly, issues include credential exposure, oversized Base64 payloads, CJK multiline shell escaping, hard-to-reverse writes, missing post-write validation, and mismatched CLI/Skill versions. This skill routes those operations through kdocs-cli: tokens are stored in the system keychain instead of chat, logs, or code comments; large content is prepared by scripts and passed as files rather than streamed Base64; complex arguments use --file or stdin to avoid encoding damage in PowerShell.

How It Works

The skill first classifies the request into create/write, partial update, type-specific features, read, file locating, file management, AI slide generation, or knowledge-base organization. It supports .otl, .xlsx, .ksheet, .pdf, .docx, .pptx, .dbt, and .form documents. A typical flow resolves the target file to file_id and drive_id, then calls write or read tools; after creation it verifies with an independent read and displays link_url or fetches the link. For unknown action, unknown service, or version-related errors, it checks CLI/Skill versions first; rate limits, circuit breakers, timeouts, and concurrency conflicts use wait, larger --timeout, serialization, or backoff.

Boundaries And Cautions

This is useful for agent workflows such as web clipping, search-read-report, table cleanup, knowledge-base archiving, and PPT generation. It is not suitable for blind bulk writes when authentication, network, or resource permissions are missing. Delete, close, cancel-share, and other irreversible operations require confirmation; batch writes should avoid lock contention; slow operations should be checked for idempotency before retry.