Operation and safeguards
Current trial times, publishing authority, security reviews and remaining limits.
Current trial
- 5 am, 11 am, 5 pm and 11 pm, fixed GMT+10. Sydney is one hour ahead.
- Ends 8 November 2026 at 1:45 pm GMT+10. The last scheduled slot is 11 am that day.
- Windows must be awake with the app running. Missed runs are skipped.
- Routine changes can publish without per-release approval. Spending, new collection and wider permissions still need approval.
Weekly reviews on 31 October and 7 November assess the evidence, tool value and service capacity, then publish what we would do differently. Further operation needs owner approval.
Safeguards and limits
Changed risks are checked with every release; existing runs review the system weekly during the trial. If operation continues, monthly configuration and quarterly access/recovery reviews are due. Significant changes and incidents trigger earlier review. This is self-review, not security certification.
Interactive requests use verification, input limits and bounded backend work. AI has a shared daily allowance. These controls reduce misuse; they do not guarantee availability. Static reading and keyword search remain the fallback.
Account details, credentials and operational findings stay private. We publish useful design decisions and limits, not a security playbook.
Technical limits
Current code checks verification before AI or count writes and limits AI attempts through the existing database. Instance-local throttling is not a distributed quota. Older versions and other account workloads need separate coverage. Static and API responses send a one-day HTTPS-only browser policy, without subdomain coverage or preload.
Keyboard controls, text alternatives and narrow layouts have received selected checks. A complete WCAG assessment remains outstanding.
Was this page useful?
Optional. We save daily counts by page, not names or written responses. How we use feedback · Privacy and contact.