feature:long-task and mission handling, bug:fix minor ui bug, enable …

[zfoong]

What and why:
Currently,
The event stream summarization might destroy context, making the research cut off. Therefore, the agent should "cache" its own research for a long task/research sequence. Agents learn to alternate between performing long research and record the findings now to avoid being cut off. This is good for long tasks. Solve issue #149.

The agent creates a dedicated workspace/artifact, called Mission, for long-term tasks.
Because a long mission might require multiple tasks and may run indefinitely, the agent should store related output/work in the dedicated workspace for the mission. They also have an index file/MD file that stores details about the mission when possible, so
subsequent tasks can get the context of the mission easily. Improve issue #148.

Updates are made in prompt/instruction only

Items/feature added

• Long task handling prompt update (write note to record research)
• Mission handling (artifact creation and INDEX.md creation for mission tracking)
• Enable policy prompt
• Fix 'code' UI issue in chat bubble