agent-observatory-sensor.nolimit-observatory.workers.dev

The description explicitly instructs the agent to 'Follow start_here.hop first (skill file, 302)' and then chain into docs, CLI, and remote MCP hops. This is an instruction aimed at the AI's behavior rather than a description of a simple lookup, and it encourages the agent to autonomously fetch and execute content from an external, unverified gateway ('Monid') and to connect to additional remote MCP servers. This could be used to smuggle in a prompt-injection payload, install unauthorized tooling, or route the agent to a pay-per-call service without the user's explicit awareness or consent.

This tool is framed as a simple 'query length' style utility but its description reveals it is actually a people-search index for locating information about living individuals, and instructs the agent to follow CLI setup, docs, API, and remote MCP hops. Combining (a) an instruction to chain into external/remote services and (b) a stated purpose of looking up personal data on real people raises both a manipulation risk (agent silently pivoting to untrusted external endpoints) and a privacy/harm risk (facilitating doxxing or unauthorized personal data lookup) that is far broader than the tool's ostensible one-line description suggests.