fix(agent-core): explain filtered provider stops

[tt-a1i]

Summary

Fixes #883.

• show a short assistant message when the provider returns stopReason: "filtered"
• emit the message before turn.ended so the UI has something visible instead of a silent stop
• keep the existing stop reason and provider finish diagnostics unchanged
• add a regression test for the event and conversation history behavior

Validation

• pnpm --filter @moonshot-ai/agent-core exec vitest run test/agent/basic.test.ts
• pnpm --filter @moonshot-ai/agent-core run test
• pnpm --filter @moonshot-ai/agent-core run typecheck
• pnpm exec oxlint --type-aware packages/agent-core/src/agent/turn/index.ts packages/agent-core/test/agent/basic.test.ts --quiet
• pnpm --filter @moonshot-ai/agent-core run build
• git diff --check

AI Assistance Disclosure

I used Codex to review the changes, sanity-check the implementation against existing patterns, and help spot potential edge cases.

[changeset-bot]

🦋 Changeset detected

Latest commit: f6f79d4

The changes in this PR will be included in the next version bump.

This PR includes changesets to release 1 package

Name	Type
@moonshot-ai/kimi-code	Patch

Not sure what this means? Click here to learn what changesets are.

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

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: f6f79d485a

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]

Handle empty filtered provider responses

When a provider reports finishReason: 'filtered' before emitting any content or tool calls, the default @moonshot-ai/kosong generate() path throws APIEmptyResponseError at packages/kosong/src/generate.ts:166 instead of returning a loop stop reason, so this new stopReason === 'filtered' branch is never reached. In that common content-filter scenario the turn still goes through the failed/error path rather than showing the intended user-facing filtered message; handle APIEmptyResponseError with finishReason === 'filtered' in the catch path or before it is converted to a failed turn.

Useful? React with 👍 / 👎.