Skill Library Chinese Name Manager

Problem

Local skill libraries often mix hand-built skills, marketplace plugins, and upstream community packages. English description fields, duplicate skills, stale versions, and MCP wrappers can make the panel hard to read. The core idea is to split high-risk writes into preview, confirmation, and rollback steps instead of overwriting local config in one pass.

How It Works and Limits

• Chinese naming: scans local skills, proposes names and summaries, and writes to _skillhub_meta.json only after confirmation or a mapping-table match.
• Library analysis: reads install directories, usage logs, brand flags, and protected entries, distinguishes installed skills from log leftovers, and presents numbered choices.
• Update and MCP checks: compares SkillHub upstream versions and recognizes http, stdio, npm, and local-script MCPs; read-only by default to avoid misreporting unpublished drafts.
• Safety boundary: dry-run is the default, and brand, built-in, connector, and self skills are excluded from rename or deletion.

It fits routine maintenance of skill names and versions, not modifying server-side connectors beyond platform permissions, and it does not guarantee that host sandboxes will not intercept deletion.