feat: hardcode tasklist GUIDs in task tool description

[lishuceo]

Summary

• 在 task 工具描述中硬编码 4 个任务清单的 GUID（Bot/MCPTools/UrhoX/Maker）
• 创建任务时 Claude 会自动根据内容选择合适的清单，不确定时默认用 Bot
• 解决了 list_tasklists API 对 bot token 返回空列表的问题

Background

list_tasklists API 使用 bot token 调用时无法返回用户创建的清单。将 GUID 直接写入工具描述，让 Claude 无需 API 查询即可将任务归入正确清单。

Test plan

• npx vitest run src/feishu/tools/__tests__/task.test.ts — 61 tests passed

🤖 Generated with Claude Code

[claude]

Claude finished @lishuceo's task in 45s —— View job

---

PR Review: Hardcode tasklist GUIDs in task tool description

• Check for previous review comments — none found
• Read changed file and understand context
• Review changes and post findings

---

✅ Approved — Clean, focused change that solves a real problem (bot token can't list user-created tasklists).

The diff is minimal and correct: 4 hardcoded GUIDs added to the tool description, tasklists changed from optional to required in the description text, and the list_tasklists reference removed from the format hint.

One observation (not blocking): The tasklists Zod schema field remains z.string().optional() (line 196), and the create handler doesn't enforce its presence programmatically (line 236). This means the "required" constraint is soft — enforced only via the LLM prompt, not at the code level. This is fine as a pragmatic workaround, but worth noting in case a future caller bypasses the LLM and calls the tool directly without tasklists.

Overall: small, well-scoped change with passing tests. LGTM.

[claude]

Claude finished @lishuceo's task in 1m 2s —— View job

---

PR Review: feat/task-tasklist-config

• Check for previous review comments — none found
• Read changed files and understand context
• Review changes and post findings

---

✅ Approved — Two well-scoped commits, both clean.

Commit 1 (29542b6): Hardcodes 4 tasklist GUIDs in the task tool description and updates the prompt to mark tasklists as required. The previous review's observation still applies — tasklists remains z.string().optional() at the Zod level (line 196) — but this is a pragmatic choice since only the LLM calls this tool.

Commit 2 (0247775): Removes session invalidation on system prompt hash change. The rationale is sound — Agent SDK's resume accepts a new systemPrompt, so the old session naturally gets updated tool descriptions. The cleanup is thorough: resumeSkipped is fully removed from types, executor return values, and the _historyDedup clearing logic. The hash computation is retained for diagnostic logging only.

Also included: maxDisplayed bumped from 16→40 in buildToolProgressCard (message-builder.ts:507). Minor UX tweak, no issues.

No issues found with confidence ≥ 75. LGTM.