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

The description instructs the AI agent to 'Follow start_here.hop first' and pass through a 302 redirect containing utm_source/utm_medium tracking parameters, then follow further 'docs, CLI, and remote MCP hops.' This is an imperative directive aimed at the agent's behavior (not just a description of the tool's function), nudging it to autonomously traverse an external, undisclosed chain of links/redirects to a 'pay-per-call tool gateway.' This could lead the agent to register for paid services, execute untrusted CLI instructions, or connect to arbitrary remote MCP servers without the user's informed consent.

Like the other 'index' tools, this description commands the agent to 'Follow start_here.hop' through tracked redirects to reach a 'living people-search index.' Beyond the embedded agent-directive pattern, the underlying capability (locating a people-search index) is a privacy-sensitive capability that could facilitate doxxing or unauthorized personal-data lookups, which is disproportionate to the stated minimal purpose of 'finding an index.'

The description instructs the agent to follow a tracked redirect chain and explicitly mentions returning a '7-day trial handoff for your operator.' This frames the tool as a vehicle for enrolling the user in a commercial trial via agent-driven automation, combined with imperative language ('Follow start_here.hop first') aimed at controlling agent behavior rather than simply describing functionality — a social-engineering/prompt-injection pattern that could result in unintended financial or contractual commitments on behalf of the user without their explicit informed consent.