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

Annotations provide readOnlyHint=false, idempotentHint=false, and destructiveHint=true, which the description does not contradict. The description adds genuinely useful behavioral context beyond the annotations: COST ($0.2/call), LATENCY (~2000ms), and especially EXECUTION: async_by_default with instruction to use get_outcome for the result. However, it never explains the irreversible/destructive aspect flagged by destructiveHint=true (what handing off permanently changes), so it doesn't fully leverage the annotation signal.

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 one-sentence core purpose is front-loaded, followed by clearly labeled sections (example, when to use, when not to use, cost, latency, execution) that each earn their place. It is longer than the minimal get_calls example, but every section adds decision-relevant information rather than filler, and the headers make it scannable.

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 moderately complex tool (5 params, nested context object, no output schema), the description covers the core action, the context bundle contents, when to use it, and how to retrieve the async result via get_outcome. Since there is no output schema, the description need not explain return values. The main gap is that the destructive implications and per-parameter semantics are not elaborated, but the overall usage picture is sufficiently complete for an agent to call this correctly.

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 only 20% (only idempotency_key has schema-level docs), so the description carries the burden for the other four parameters. The example call demonstrates smb_id, reason, context, and priority usage in a realistic structure, which is helpful, but it does not explain the meaning of each parameter beyond what the schema enumerations already show. The context object's nested fields are left entirely to the schema. The example partially compensates for the coverage gap but not fully.

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 states a specific verb ('Hand off') and resource ('an in-flight task to a human operator'), plus the makeup of the context bundle (transcript, prior actions, identifiers, recommended next step). This clearly distinguishes escalation from siblings like call_business or send_message, which perform automated actions. It stops short of explicitly naming a sibling to differentiate against, so it doesn't reach a 5.

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 WHEN TO USE / WHEN NOT TO USE sections give explicit conditions ('after channel-fallback exhaustion', 'requires human judgment', 'customer has explicitly requested human contact') and an explicit exclusion ('Do not use as a first resort... only after automated resolution attempts'). This is strong guidance, though it frames alternatives by condition rather than by naming specific sibling tools, so it falls just below the top tier.

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