new·The score now tells you which way it movedA brain's exam only ever grows: its own material writes questions, and so does every question a real caller asked and did not get answered. The score is a percentage over that growing set, so a brain that learned more could post a smaller number — and this week three did. One of them answered two MORE questions than the week before and showed eighteen points less. Printed as a single percentage, that reads as decline to a reader and as punishment to anyone who contributes material.all news →
mozg.beta
Sign in

OWASP Cheat Sheets · all subjects

logging/mcp_servers

3 notes, read out of this brain and free to use. Each one was extracted from a source and is re-checked against its exam.

mcp_prompt_injection event logging

Log prompt injection attempts in MCP clients/servers with event name 'mcp_prompt_injection' followed by userid. Level is WARN. Indicators include instructions to ignore system messages, attempts to override tool policies, requests to reveal secrets, or attempts to coerce tool calls outside intended task. Avoid logging full prompt/tool I/O; prefer logging detection category or rule ID.

mcp_resource_exhaustion event logging

Log resource exhaustion attempts in MCP clients/servers with event name 'mcp_resource_exhaustion' followed by userid. Level is WARN. Examples include unusually large prompts, repeated retries, tool-call loops, or behavior driving excessive token usage/cost. Avoid logging full prompts; prefer logging measured usage (tokens in/out), policy thresholds, request IDs, and tools/models involved.

mcp_tool_poisoning event logging

Log detection of malicious or tampered tools in MCP with event name 'mcp_tool_poisoning' followed by userid. Level is WARN. Examples include unexpected tool changes, signature/hash verification failures, suspicious metadata, or tools from unvalidated marketplaces. Avoid logging full tool details; prefer tool name, source, version, signature status, and policy applied.

Give your agent this brain