Return memories known to have failed — invalidated, superseded, or
AI agents call failed_hypotheses to retrieve information from State Trace without modifying anything — typically the context-gathering step in research, monitoring, and reporting workflows, before the agent takes action elsewhere.
This tool retrieves or queries existing data (failed hypotheses from working memory) without side effects. It performs read-only access to the agent's causal memory system, similar to a get or fetch operation. There is no indication it modifies, deletes, or executes external operations.
From the tool's definition Tool name 'failed_hypotheses' and description indicating it 'Return[s] memories known to have failed' — a retrieval operation with no modification, creation, or deletion of data.
Attacks that exploit this kind of access
Return memories known to have failed — invalidated, superseded, or. It is categorised as a Read tool in the State Trace MCP Server, which means it retrieves data without modifying state.
Register the State Trace MCP server in PolicyLayer and add a rule for failed_hypotheses: allow, deny, rate-limit, or require approval. Point your MCP client at the PolicyLayer proxy URL and the rule is enforced on every call, before it reaches State Trace. Nothing to install.
failed_hypotheses is a Read tool with low risk. Read-only tools are generally safe to allow by default.
Yes. Add a rate_limit block to the failed_hypotheses rule in your PolicyLayer policy. For example, setting max: 10 and window: 60 limits the tool to 10 calls per minute. Rate limits are tracked per agent session and reset automatically.
Set action: deny in the PolicyLayer policy for failed_hypotheses. The AI agent will receive a policy violation error and cannot call the tool. You can also include a reason field to explain why the tool is blocked.
failed_hypotheses is provided by the State Trace MCP server (agent-pattern-labs/state-trace). PolicyLayer sits as a proxy in front of this server to enforce policies before tool calls reach the server.
Every MCP server has a record like this.
Type a name, get the same breakdown: verified identity, auth posture, risk grade, capabilities, recommended policy.
Teams ship this data inside their own products. See what a licence covers →