廉价的存活/进度快照(比 agent_read 轻):running/settled/nSteps/checkpointSeq/错误/contentTail 短尾/driftHint。
AI agents call agent_state to retrieve information from Frx Director without modifying anything — typically the context-gathering step in research, monitoring, and reporting workflows, before the agent takes action elsewhere.
This tool fetches a lightweight snapshot of the agent's current state (running status, progress steps, checkpoint sequence, errors, content tail, drift hint). It is a read-only status query with no side effects, similar to a polling/health-check endpoint. Severity is low since misuse only exposes agent progress metadata.
From the tool's definition 存活/进度快照 (alive/progress snapshot), 'running/settled/nSteps/checkpointSeq/错误/contentTail 短尾/driftHint' — purely retrieves status information with no side effects
Attacks that exploit this kind of access
廉价的存活/进度快照(比 agent_read 轻):running/settled/nSteps/checkpointSeq/错误/contentTail 短尾/driftHint。. It is categorised as a Read tool in the Frx Director MCP Server, which means it retrieves data without modifying state.
Register the Frx Director MCP server in PolicyLayer and add a rule for agent_state: 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 Frx Director. Nothing to install.
agent_state 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 agent_state 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 agent_state. 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.
agent_state is provided by the Frx Director MCP server (whitenightshadow/frx-director-mcp). 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 →