WritersLogic verifiable AI agents

llm-watermark-detector

Detect LLM text watermark evidence: green-list z-test for a key you hold, plus invisible-character markers — Tests whether text carries watermark evidence, for a pipeline that must decide whether to trust or label content. Two independent checks: (1) a green-list z-test (Kirchenbauer et al. 2023) over green flags you supply, or over token ids and a key using this endpoint's own HMAC-SHA256 scheme, returning the z-score and exact binomial p-value; (2) an invisible-character scan of the text for zero-width

Tool call template

This tool declares no required parameters. Add optional parameters from the input schema as needed.

{
  "id": 1,
  "jsonrpc": "2.0",
  "method": "tools/call",
  "params": {
    "arguments": {},
    "name": "llm-watermark-detector"
  }
}

Input Schema

Name	Required	Description
`key`	No	Secret key for tokenIds. Used for this call only and never stored or echoed.
`text`	No	Text to scan for invisible markers.
`gamma`	No	Expected green fraction under no watermark. Default 0.25.
`scheme`	No	Scheme for tokenIds; default and only value.
`tokenIds`	No	Token ids in order (at most 5000). Green membership is computed from the key with the hmac-sha256-green-1 scheme; each distinct consecutive pair is scored once.
`minTokens`	No	Fewer scored tokens than this is unknown. Default 25.
`greenFlags`	No	One 0/1 per scored token: 1 when your own detector's green list contains the token. Deduplicate repeated n-grams yourself if your scheme requires it.
`zThreshold`	No	z at or above which the result is fail. Default 4 (Kirchenbauer et al.).

Schema Changelog

Changes observed during successful MCP inspections.

First observedNov 15, 2026

TDQS

A4/5.0

Behavior3/5

Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?

With no annotations at all, the description carries the full behavioral burden. It does add real method transparency — Kirchenbauer et al. attribution, the statistic used, and the exact return values ('z-score and exact binomial p-value') — but leaves out failure/edge behavior (what a below-minTokens 'unknown' looks like, what happens with no zThreshold) and cuts off mid-phrase at 'for zero-width', so the second check's behavior is never actually described.

Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.

Conciseness3/5

Is the description appropriately sized, front-loaded, and free of redundancy?

It is front-loaded correctly, leading with the verb and the two-mode split. However the prose is a dash-chained run-on and the final clause is truncated at 'invisible-character scan of the text for zero-width', leaving a sentence that never finishes and a dangling, uninformative fragment.

Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.

Completeness4/5

Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?

Eight parameters with zero required fields and no output schema is a demanding case, and the description does cover both input routes and the primary statistic returned. The gaps are the truncated second check (no statement of what the invisible-character scan returns) and no clarification of which parameter combinations are valid calls.

Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.

Parameters4/5

Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?

Schema coverage is 100%, so a baseline of 3 is warranted, but the description genuinely adds semantic structure the schema does not: it frames greenFlags and tokenIds/key as two mutually exclusive routes to the same z-test and notes that both feed the same scoring. It does not explain the gamma/zThreshold/minTokens trade-offs beyond what the schema already says.

Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.

Purpose5/5

Does the description clearly state what the tool does and how it differs from similar tools?

The description opens with a concrete verb+resource ('Detect LLM text watermark evidence') and immediately decomposes the tool into two named, independently scoped checks (green-list z-test; invisible-character scan). The mode carve-out ('green flags you supply, or over token ids and a key') makes the tool's scope unambiguous even though no sibling is named.

Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.

Usage Guidelines4/5

Does the description explain when to use this tool, when not to, or what alternatives exist?

It explains the decision context well ('for a pipeline that must decide whether to trust or label content') and gives intra-tool routing between the two input styles (your own green flags vs. tokenIds + key). It does not state when *not* to use the tool, any prerequisite (e.g. minimum text length, that you must already hold a key for the HMAC path), or name an alternative.

Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.

All WritersLogic verifiable AI agents tools Inspect

Try in Browser

One endpoint for every MCP toolRayrun runs every MCP server behind one URL and holds the credentials. Each call is decided per client, tool, and argument, then recorded.ray.runAd

Glama MCP Gateway

Add one secure layer between your agents and this server.

Resources

Need Help?