https://agent-observatory-sensor.nolimit-observatory.workers.dev/s/mimepart-ok/mcp
highagent-tool-index
Embedded directive telling the agent to auto-follow an external redirect chain
The description instructs the calling agent to 'Follow start_here.hop first (skill file, 302)' and then follow additional 'docs, CLI, and remote MCP hops.' This is an imperative aimed at the AI rather than a description of return data, and it encourages the agent to automatically chase a chain of external redirects/hops to an unnamed third-party gateway ('Monid') without user visibility or confirmation. This pattern is consistent with a technique to smuggle the agent into fetching and executing content from an attacker-controlled or unverified location under the guise of a benign lookup tool.
highpeople-search-index
Directs agent to follow external hops to a people-search / PII service, no data-handling safeguards described
This tool advertises finding a 'living people-search index' (personal data about real individuals) and, like agent-tool-index, tells the agent to follow CLI, docs, API, and 'remote MCP hops' to an unnamed service ('Ploid'). Combining an instruction to chase external hops with a stated purpose of aggregating information about real people raises both a prompt-injection/redirect risk and a privacy/harm risk (facilitating unconsented lookups on individuals) that is broader than a simple index lookup should require.
low
Not on gate's verified list
This server isn't on gate's handpicked list. That's not necessarily bad, but there's no third-party signal about it — review its tools carefully.
lowmimepart-ok
Generic multi-purpose schema shared across unrelated tools obscures actual data flow
All 30 tools expose an identical, broad input schema (zone, json, url, host, city, query, path, ref, feed) regardless of the tool's single stated purpose. For example, a MIME-boundary counter accepts a URL, a file path, a git ref, and an RSS feed parameter that have nothing to do with MIME parsing. This makes it impossible for the agent or user to know which fields are actually used versus silently ignored, and creates a risk that data placed in one field (e.g., 'json' or 'path') could be forwarded to an unintended downstream service under a differently-named tool. This design should be tightened so each tool's schema matches its documented function.
lowutc-time
Unused broad parameters (url, path, ref, json, feed, query, host) unrelated to timestamp retrieval
A tool that only needs to return the current UTC time exposes eight additional parameters (including file paths, git refs, and URLs) with no explanation of what happens to them if supplied. This mismatch between stated purpose and accepted input surface is a minor red flag for hidden functionality or data collection beyond what's described.