docs: require user-facing changeset wording, rewrite the v1-mcp entry

[bj456736]

Motivation

Recently merged changesets show two recurring wording problems:

• .changeset/v1-mcp-management-plane.md shipped an SDK paragraph that reads like an API reference — a ten-line enumeration of method names (listMcpServers, getMcpServer, testMcpServerConfig, addMcpServer, reconnectMcpServer, listMcpServerAuthStatuses, inspectAppMcpServers, …). The current skill explicitly allows this ("write those entries honestly and technically; their wording does not need to suit end users"), so the skill is the right place to fix it.
• New feature entries give no maturity/availability signal — e.g. the /tower entry does not say whether the feature is experimental, gated, or generally available, so readers cannot tell what they can actually use.

Changes

.agents/skills/gen-changesets/SKILL.md

• Reword the SDK/internal-package license: nobody curates those changelogs later, so entries must still be short and readable — summarize the capability change (what callers can now do, what behavior changed) in one or two sentences; never enumerate methods/parameters/fields like an API reference.
• New wording rule: write for a reader who does not know the codebase; one outcome per entry; no semicolon/slash-list sentence stuffing — split into separate changesets instead.
• New wording rule: say who can actually use the change — experimental, flag-gated, engine-gated (legacy / v2), platform- or surface-specific (TUI / kimi web / VSCode) must be stated in plain words.
• New workflow step: re-read the finished entry from a user's perspective before committing; for user-facing changes the human author should proofread the entry before merge.
• Matching red flags for all of the above.

.changeset/v1-mcp-management-plane.md

Rewrite the entry per the new rules. Frontmatter unchanged. The CLI paragraph keeps its three effects but is split into two readable sentences; the SDK paragraph becomes a capability-level summary with zero method-name enumeration.

Before (SDK paragraph, abridged): "the MCP management surface is now backed by a unified, source-tagged registry — listMcpServers also covers plugin-declared servers … new getMcpServer … testMcpServerConfig … addMcpServer … reconnectMcpServer … listMcpServerAuthStatuses … inspectAppMcpServers … authenticateAppMcpServer / resetAppMcpServerAuth …" (~10 lines)

After: "MCP servers are now managed through a unified registry that also covers plugin-declared servers. SDK callers can list, inspect, probe, add, reconnect, and authenticate MCP servers at runtime — including plugin servers — with effective config and probe-verified authorization state. Stored OAuth grants now record expiry and refresh proactively, concurrent logins for the same credential share one browser flow, and secret-bearing env / headers values are redacted to key lists in status responses."

No code changes; no changeset needed for this PR itself.

[changeset-bot]

⚠️ No Changeset found

Latest commit: 1f3a176

Merging this PR will not cause a version bump for any packages. If these changes should not result in a new version, you're good to go. If these changes should result in a version bump, you need to add a changeset.

This PR includes changesets to release 4 packages

Name	Type
@moonshot-ai/kimi-code	Patch
@moonshot-ai/kimi-code-sdk	Minor
kimi-code	Patch
@moonshot-ai/acp-adapter	Patch

Click here to learn what changesets are, and how to add one.

Click here if you're a maintainer who wants to add a changeset to this PR

[pkg-pr-new]

pnpm dlx https://pkg.pr.new/@moonshot-ai/kimi-code@1f3a176

npx https://pkg.pr.new/@moonshot-ai/kimi-code@1f3a176

commit: 1f3a176

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 1f3a176741

ℹ️ About Codex in GitHub

Codex has been enabled to automatically review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".

[chatgpt-codex-connector]

Restore the plugin scope in the legacy-engine summary

The listed install, enable/disable, remove, and reload actions are plugin-management operations: the corresponding KimiCore plugin mutators are what call syncPluginMcpServersInSessions. Removing “plugin” from the previous wording makes this read as though MCP servers themselves expose all five operations and obscures which user workflow gained live-session synchronization; retain the plugin qualifier or explicitly describe plugin-provided MCP servers.

Useful? React with 👍 / 👎.

[chatgpt-codex-connector]

Apply the new concise-entry rule to the rewritten changeset

This paragraph still combines the registry change, six runtime-management capabilities, OAuth expiry/refresh behavior, login deduplication, and secret redaction into three dense sentences. Because this raw text is copied directly into the SDK changelog, it remains exactly the multi-outcome technical paragraph that the newly added “one outcome per entry” and concise-wording guidance is intended to prevent; split the independent outcomes or retain only the primary caller-facing capability.

Useful? React with 👍 / 👎.