Refine daily agent: better signal from chats and activity

[unforced]

Context

The daily reflection agent has access to journal entries and chat logs, but the output isn't as useful as it could be. This is partly a system issue (what data is available, in what shape) and partly a personal agent-tuning issue (prompt, what to focus on).

Noting that a lot of the value here comes from personal customization — this issue tracks the system-level improvements that make personal agents better for everyone.

System-level improvements

Better activity data

• The activity hook already writes structured data to Daily/.activity/{date}.jsonl — consider adding a read_activity_log tool that gives agents clean access to this
• Activity log has: timestamp, session_id, session_title, agent_type, exchange_number, summary — this is good signal for "what did I actually do today"

Chat tool quality

• Audit what read_chat_log and read_recent_sessions actually return — is it useful or noisy?
• Consider a tool that returns chat summaries rather than full transcripts
• Balance: agents need enough context to make connections, but not so much they get lost

Tool gaps

• Can agents read brain entities? If the brain has context about projects, that's useful for daily reflection
• Can agents see what cards other agents produced? Useful for building on each other's work

Personal scope (testbed, not universal)

• Prompt refinement: focus on work accomplished, project connections, open threads
• This is the testbed for how users customize their own agents
• Learnings here inform the defaults and documentation

Depends on

• Generalize agents issue (process-day pattern)
• Card types (process-day producing multiple card types)