Feature request: expose timestamps to Claude as structured data for time-aware reasoning

Problem

Claude Code has no concept of time within a session. It cannot:

• Know how long it's been since the last message or tool call
• Correlate its own actions with external systems by time
• Calculate durations between its own events
• Detect stale state ("this agent has been running for 20 minutes, might be hung")
• Make time-aware decisions ("5 minutes since last CI check, time to poll")
• Build accurate timelines in session summaries

Currently the only workaround is shelling out to date, which is hacky and only gives "now" — not when past messages/results arrived.

Proposed solution

Expose timestamps as first-class data Claude can read and reason with:

• Each message and tool result should carry a timestamp accessible to Claude
• A lightweight way to get "now" without date shell calls
• Durations between events should be computable
• Background task notifications should include start time, not just completion time

Why this matters

Claude increasingly operates as a long-running orchestrator — managing parallel agents, polling CI, waiting on background tasks. Without time awareness it's flying blind on the temporal dimension. Time is a basic data point that every other development tool exposes.

Example use cases

• "The benchmark started at 10:42 and finished at 11:18 — 36 minutes for 120 calls"
• "Agent X has been running for 25 minutes vs Agent Y which finished in 3 — X might be stuck"
• "Last CI poll was 8 minutes ago, checking again"
• Session summaries with accurate timing, not guesses