I filed a critical bug against my own tool. Then I read the code — and my own root cause was wrong.

Short summary

A developer filed a critical bug against their own MCP server (safari-mcp) where an AI agent navigated user-owned browser tabs. The proposed fix—injecting a sentinel marker for tab identity—had already shipped months earlier. The real bug was a clamping fallback four lines below the correct identity-resolution code that silently redirected the agent to the last tab when the tracked index went out of range.

• •Author filed a serious bug in safari-mcp where agent touched user tabs, breaking the core safety promise
• •Proposed sentinel-based fix was already implemented; real root cause was an index-clamping fallback that targeted the wrong tab
• •Lesson: reasoning from architectural memory without reading code leads to confidently wrong root-cause analysis

Generated with AI, which can make mistakes.