Soft-delete the latest HTML version for a deal. The /browse
AI agents call html_reset to permanently remove resources in Llama — typically in cleanup and lifecycle workflows. It does its job in a single call, and there is no undo.
Destructive operations that delete data, even if technically recoverable through archives or backups, are classified as Destructive rather than Write. The tool removes content irreversibly from the user's perspective. The context of deal management on a venture platform means this affects business-critical documents.
From the tool's definition 'Soft-delete the latest HTML version for a deal' — removes data from a deal record. While described as 'soft-delete', it permanently removes the latest HTML version and cannot be undone within normal operations.
Attacks that exploit this kind of access
Soft-delete the latest HTML version for a deal. The /browse. It is categorised as a Destructive tool in the Llama MCP Server, which means it can permanently delete or destroy data. Block by default and require explicit approval.
Register the Llama MCP server in PolicyLayer and add a rule for html_reset: 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.
html_reset is a Destructive tool with critical risk. Critical-risk tools should be blocked by default and only enabled with explicit human approval.
Yes. Add a rate_limit block to the html_reset 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 html_reset. 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.
html_reset 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.
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 →