What This Looks Like
The user asks for a bounded action, but the AI adds extra steps, edits, checks, recommendations, files, changes, or follow-up tasks. The added work may look helpful on the surface, but it was not requested and may change the scope, output, cost, or review burden of the task.
Why It Matters
Extra work can create hidden obligations. A user may now have to review changes they did not ask for, undo unrelated edits, sort required work from optional work, or determine whether the AI changed the goal. In agentic workflows, unsolicited work can also trigger tools, modify state, or affect downstream steps without clear authorization.
Structural Signal
The requested task has a declared scope, but the system performs or proposes work outside that scope without a sufficient authority boundary. The issue is not that the AI made a helpful suggestion; it is that optional or undeclared work becomes part of the executed task.
Common Triggers
- The prompt leaves the boundary between requested work and optional help unclear
- The system treats adjacent improvements as authorized work
- The agent is rewarded for initiative without a stop rule
- The task lacks a declared allowed-change surface
- Follow-up recommendations are mixed into execution
- The AI interprets a narrow request as permission to optimize the broader situation
When to Use This Issue
Use this Issue when the AI or agent adds work that was not part of the user’s request and that extra work changes the task surface, review burden, workflow state, or expected output.
When Not to Use This Issue
Do not use this Issue when the AI asks a clarifying question, offers optional suggestions clearly separated from the requested work, or adds a small necessary step required to complete the task. Use this Issue when the added work crosses the requested boundary.