fix(cursor): use plan.limit for usage calculation, not breakdown.total

[bcharleson]

Problem

Cursor usage was always showing 100% used (0% left) regardless of actual usage.

Root Cause

The code was using breakdown.total as the denominator for calculating usage percentage, but breakdown.total represents the used amount (included + bonus credits consumed), not the plan limit.

This caused the calculation to be used/used = 100% instead of used/limit = actual%.

API Response Structure

"plan": {
  "used": 3035,        // credits consumed (cents)
  "limit": 40000,      // actual plan allowance (cents) ← should use this
  "breakdown": {
    "included": 3035,  // included credits used
    "bonus": 0,        // bonus credits used  
    "total": 3035      // total USED (included + bonus), NOT the limit!
  }
}

Fix

Changed to use plan.limit directly instead of breakdown.total ?? plan.limit.

Testing

• Verified with CLI: now shows correct ~8% used instead of 100%
• Verified in app UI: shows 92% left correctly

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 5179e60a26

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[chatgpt-codex-connector]

Ensure daily keepalive actually fires

With daily mode, SessionKeepaliveManager.calculateCheckInterval polls on a fixed 3600‑second cadence, while shouldRefreshSession only triggers when the current time is within 5 minutes of the scheduled hour/minute. If the task starts at a minute offset more than 5 minutes away from the target time, each hourly check will repeat that offset and never enter the 5‑minute window (e.g., start at 10:10 for a 10:00 schedule), so the daily refresh never runs. Consider computing the next scheduled fire time (and sleeping until then) or polling more frequently so the scheduled window is guaranteed to be hit.

Useful? React with 👍 / 👎.