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

Annotations declare readOnlyHint=false, openWorldHint=true and destructiveHint=false; the description usefully adds that this actively sends real attack traffic to the endpoint and returns a pass/block breakdown, which goes beyond the annotations. It stops short of saying whether an eval record is persisted, whether the run is synchronous or must be polled, or roughly how long 39 prompt attacks take — all material for a non-destructive but open-world outbound call.

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?

Two sentences, front-loaded with the action and then the mechanism and return shape. Every clause carries information — prompt count, attack taxonomy, and result semantics — with no filler or repetition of the title.

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?

For a 5-parameter, no-output-schema tool, the description covers what it does and roughly what it returns, which is adequate. The notable gap is lifecycle: it never says whether results are returned inline or come back later via get_agent_eval_results, nor does it mention the required `context` reporting parameter, leaving the agent unsure how to complete the workflow.

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 coverage is 100%, so all five parameters are already documented, establishing baseline 3. The description's enumeration of the six attack categories usefully mirrors the valid values for the `categories` filter and reinforces what `severity_min` gates, but it adds no format, default, or interaction detail beyond the schema.

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?

Specific verb and resource ("Run adversarial safety testing against an AI agent") plus a concrete scope of 39 attack prompts across six named categories and a stated outcome (which succeeded vs were blocked). An agent can tell this is an active security probe, not a report fetch. It does not, however, distinguish itself from the closely-named sibling run_agent_eval, so the agent must guess which runner to pick.

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?

Usage is implied by "against an AI agent" and the required agent_endpoint, but there is no explicit when-to-use, no prerequisites (e.g., that the target must be reachable), and no exclusions. It also never routes the agent to related siblings such as list_adversarial_prompts (to preview the attacks) or get_agent_eval_results (to retrieve results), leaving the workflow to inference.

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