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

Annotations already declare readOnlyHint=true and destructiveHint=false, so the description adds significant behavioral context: it notes that the tool is a 'heuristic aid', not a security boundary, and explains its limitations ('a 'low' verdict is not evidence that a tool is safe, and an injection phrased to avoid the patterns will score low'). It also states it is read-only, analyzes only the passed text, and requires a Kamy API key. This goes well beyond the annotations, providing essential transparency about the tool's fallibility and operational constraints.

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

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

The description is a single, well-structured paragraph of about 100 words. It front-loads the purpose and key differentiators, then covers usage guidelines and behavioral notes. Every sentence adds distinct information: purpose, use case, output type, limitations, and authentication requirement. No redundancy or filler.

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

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

Given that the tool has 2 parameters, no output schema, and no nested objects, the description is richly complete: it explains what the tool does, how to use it, what to expect as output (risk levels and excerpts), its limitations, and the auth requirement. This is more than sufficient for a tool of this complexity, especially with the complementary annotations and schema coverage.

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

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

Schema description coverage is 100%, and the description adds context for both parameters: it explains that 'description' is for a single tool description to scan, while 'tools' is for multiple assets (e.g., from a tools/list result), and that either supply is acceptable. The description also clarifies that the tool expects description text specifically for prompt injection scanning, not arbitrary input. Despite the schema already documenting the fields, the description elaborates on the use cases and input types, adding value.

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

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

The description clearly states that the tool scans MCP tool description text for prompt-injection tells, specifying the verb 'scan', the resource ('tool description'), and the scope ('instructions addressed at the reading model, data-exfiltration hints, attempts to override your system prompt or hide content'). It also explains the output (risk level and findings), fully differentiating it from sibling tools like 'ask_kamy', 'attest_artifact', or 'verify_mcp_server', which have different purposes.

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

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

The description explicitly states when to use the tool ('Run it on descriptions from third-party MCP servers before you act on what they say'), what it should not be treated as ('a security boundary', not 'clearance to trust an untrusted tool'), and what the model should do instead ('keep your own judgement and human review in the loop'). This provides comprehensive guidance, including exclusions and alternatives.

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