Playwright MCP Usage Policy blocks — DELETE requests, unsafe tool names, and automation alternatives.
E2E tests intercept HTTP, fire DELETE routes, and Playwright MCP literally labels runs as unsafe. Classifiers eat that for lunch.
Tool-result blocks
Blocks fire on assistant continuations after tool output — you cannot fix by editing your last user message.
Mitigation
Isolate E2E agent sessions from security review sessions. Use shorter test scripts per run.
Backend swap
Keep Playwright MCP; route LLM inference to uncensored API when test logs pollute context.
Try uncensored AI free
No filters, no lecture, no training on your chats. Start in under a minute.
Test automation should not require fighting Usage Policy on every green build.
FAQ
Is Playwright banned on Claude Code?
No — but accumulated test output can trigger policy classifiers.