Understanding Scope + Safety
What scope drift events, safe fix actions, and approvals avoided tell you about your team's AI coding discipline.
What this report tracks
The Scope + Safety report shows how well your team's AI runs stayed within their declared scope. It captures every event where Avorelo intervened to enforce a boundary, apply a safe fix, or flag something that needed attention before the run could proceed.
The report is designed for team leads and engineering managers who want to understand whether scope discipline is holding up across the team as AI task frequency increases.
Three types of events
Contained
A scope boundary was enforced before any out-of-scope file was written. The agent's planned changes fell outside the declared boundary, Avorelo caught it, and the run continued with corrected scope. No escalation required.
Repaired
A safe fix was applied automatically. Unused import, whitespace normalization, or another low-risk cleanup that met the configured safe fix criteria. Applied without escalation and recorded in the receipt.
Flagged
Something was detected that did not meet the criteria for automatic repair, and the run could not confirm a clean exit. This appears in the task list as Review needed and requires human attention before the output is treated as done.
Files kept out of scope
This is the top-level metric: how many files were excluded from writes this period because they were outside the declared task boundary. A higher number means more active enforcement, not necessarily more problems.
Compare this number against the total file change count to understand the ratio of enforced exclusions to actual changes. A team doing a lot of complex refactoring will see this number increase naturally.
Approvals avoided
Each time a task stays within its declared scope and exits clean, it avoids a potential approval escalation. This is the most operationally significant metric in the report for teams that are trying to maintain delivery pace without introducing new review bottlenecks.
Scope + Safety is not a trust report. It is a discipline report. Higher containment events with low flagged events mean the system is working. Frequent flagged events may indicate tasks that need clearer scope declarations at the start.
Per-member scope adherence
The bar chart shows each member's clean exit rate: what proportion of their runs stayed within the declared scope without a flagged event. A member with a lower rate is not necessarily doing worse work. They may be running more exploratory or high-risk tasks that naturally produce more containment events.
Use this view to start a conversation, not to draw conclusions. Compare the scope adherence chart against the task list filtered by that member to understand context.
Open Scope + Safety report
View scope events and safety actions for your team