[Astro-Han]

Summary

The 0.2.2 source-candidate workflow fails during extracted-source tests because the memory-network suite packages the Feishu example from scripts/plugins, which is deliberately excluded from ASF source archives. Normal checkout CI passes because that example is present.

Move that integration test unchanged into its own test file and exclude it alongside its repository-only dependency. Also exclude the proactive assistant source-loop integration test, which imports the same example's cache, synchronization, and scheduler modules. The existing assistant runner still discovers and runs both tests in checkout CI. Keep the other external-source memory tests in the archive, and extend the existing archive regression to verify both sides of this boundary. Product behavior and release contents outside these test exclusions are unchanged.

Refs #6148. Failed qualification: https://github.com/apache/maka/actions/runs/39192269771

Verification

• Both checkout test files: 7 tests passed, including the real packaged Feishu integration.
• Complete checkout assistant suite: 182 tests passed (80 memory-network, 24 index-initiative, 78 proactive-matters), including both Feishu integration tests.
• Actual candidate created from 4e3ef5211 and extracted: all 180 retained assistant tests passed (79 memory-network, 24 index-initiative, 77 proactive-matters), with scripts/plugins and both dependent integration tests absent. This local check reused installed dependencies and their built workspace outputs; the full clean-install qualification remains with the source-candidate workflow after merge.
• Source-release suite: 23 tests passed. The new archive-boundary assertion failed before the export rule and passed after it.
• Removing the proactive source-loop export rule also causes the extended archive regression to fail.
• npm run format, npm run lint, and git diff --check passed.
• No product code changed. Core, storage, MCP, and runtime builds passed to prepare the assistant checks; full local builds, typecheck, and E2E were not rerun.

AI use

• No generative tool made a substantive contribution
• Generative tooling made a substantive contribution

Tool(s) and scope: Codex diagnosed the qualification failure, split the test, updated archive coverage, and ran the focused checks.

Checklist

• Tests cover the change and fail without it
• Lint, format, typecheck and the affected suites pass locally

Does this PR entail a change in behavior?

• Yes — described under Summary above
• No

[Sun-GLiang]

The test extraction preserves the original logic, and both memory-network source test files pass locally (7/7). However, another Feishu-dependent assistant test remains in the source archive and is still discovered by the extracted-source test runner. Please address the remaining dependency described inline.

Validation: inspected git archive output at this commit. Source-release suite: 22/23 passed; the remaining test could not complete GPG key generation because gpg is unavailable locally. Remote test/package checks were still running.

[Sun-GLiang]

Reviewed the follow-up commit 4e3ef52. The previously reported missing exclusion is fixed, and the archive regression now covers the proactive source-loop integration test.

Verified the actual git archive at this revision: both repository-only Feishu integration tests are excluded, generic memory source tests remain, and archived assistant test sources contain no remaining references to the Feishu example or scripts/plugins. Both integration files remain in the checkout and are still discovered by the existing runner. The follow-up changes only export attributes and regression coverage; no additional issues found.

Approval is based on code review and local archive verification; CI completion was not required for this review.

[M4n5ter]

English

Looks good to me on 4e3ef5211. I didn’t find any correctness or maintainability issues in this change.

The moved test preserves the original behavior, and the checkout runner still discovers both integrations. I checked the actual source archive: only the two intended test files are excluded from the assistant suites, and the remaining source tests are retained unchanged. The precise export rules and the archive assertions cover both sides of that boundary.

All 23 source-release tests passed locally, and I created, verified and reproduced the source archive from this revision. I haven’t rerun the complete assistant suite or the extracted-source qualification with a fresh install locally.

中文

看过 4e3ef5211，这次改动没有发现正确性或可维护性问题。

搬迁的测试保留了原有行为，checkout runner 仍会发现并执行这两项集成测试。我也检查了实际源码归档：助手测试中仅排除了预期的两个文件，其他来源测试保持不变。精确的导出规则和归档断言同时保护了排除与保留两侧的边界。

本地 23 个源码发布测试全部通过，也从当前提交生成、校验并重建了源码归档。没有在本地重跑完整助手套件，或通过全新安装完成解压后源码验收。