record_eval_feedback · Llama MCP: Risk & Policy

WHAT IT DOES

What record_eval_feedback does on Llama

AI agents use record_eval_feedback to create or update resources in Llama — usually the action step of a workflow, after the agent has gathered context. Every call changes real data in your Llama environment.

WHY IT NEEDS A POLICY

Why record_eval_feedback needs a policy

This tool writes feedback data associated with query results. While it does not delete data (which would be Destructive) nor execute arbitrary code (which would be Execute), it creates or modifies feedback records. The severity is medium rather than high because feedback marking typically has limited blast radius and does not directly affect financial transactions or irreversible data loss.

From the tool's definition 'Mark the latest llama CLI/MCP result as good/bad or add a real query' — the tool records feedback and modifies state by marking results, which constitutes data modification rather than pure retrieval.

Attacks that exploit this kind of access

Recommended rule Ready

Verdict

record_eval_feedback Rate-limited

record_eval_feedback stays usable, but capped — an agent stuck in a loop can't make hundreds of changes a minute. Everything else on the server is denied unless you say otherwise.

View as policy code

policy.json

{
  "version": "1",
  "default": "deny",
  "tools": {
    "record_eval_feedback": {
      "limits": [
        {
          "counter": "record_eval_feedback_rate",
          "window": "minute",
          "max": 30,
          "scope": "grant"
        }
      ]
    }
  }
}

LIMIT THIS TOOL →

Instant setup, no code required.

PolicyLayer is an MCP gateway: every tool call is checked against your rule before it reaches Llama, so you can run agents against it safely.

FAQ

Questions about record_eval_feedback

What does the record_eval_feedback tool do? +

Mark the latest llama CLI/MCP result as good/bad or add a real query. It is categorised as a Write tool in the Llama MCP Server, which means it can create or modify data. Consider rate limits to prevent runaway writes.

How do I enforce a policy on record_eval_feedback? +

Register the Llama MCP server in PolicyLayer and add a rule for record_eval_feedback: 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 Llama. Nothing to install.

What risk level is record_eval_feedback? +

record_eval_feedback is a Write tool with medium risk. Write tools should be rate-limited to prevent accidental bulk modifications.

Can I rate-limit record_eval_feedback? +

Yes. Add a rate_limit block to the record_eval_feedback 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.

How do I block record_eval_feedback completely? +

Set action: deny in the PolicyLayer policy for record_eval_feedback. 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.

What MCP server provides record_eval_feedback? +

record_eval_feedback is provided by the Llama MCP server (llama-ventures/llama-cli). PolicyLayer sits as a proxy in front of this server to enforce policies before tool calls reach the server.