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

Annotations only supply the safety profile (readOnlyHint=false, destructiveHint=false, openWorldHint=false); the description goes well beyond them by disclosing the 4-watch cap, webhook rate limit (one event per 2 s, 16 watches per account), timeout behavior (1 h default, 0 = until stopped), and the key operational fact that reading never wakes a paused sandbox.

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?

Dense, front-loaded and organized by action with semicolon-separated clauses, so no sentence is wasted. It is information-rich rather than bloated, though the action list runs long enough that an agent must parse several clauses to find the one it needs.

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 15-param mutation-capable tool with no output schema, the description supplies the return contract for read (events since cursor, folded writes with count, nextCursor) and the lifecycle constraints. It leaves the semantics of batchMs, maxWatches and idempotencyKey to the schema, but overall an agent has enough to call it 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?

With 67% schema coverage across 15 params, the description carries real weight by mapping arguments to actions (path/recursive/events/include/exclude/timeoutMs/webhook for start; watchId/cursor/waitMs for read) and explaining folded write counts. It omits batchMs, maxWatches and idempotencyKey, which only the schema documents, but the core semantics are covered.

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+resource ('Watch a sandbox file or directory for changes') and then enumerates the four sub-actions (start/read/list/stop) with their arguments and return shapes. An agent can distinguish this watch-based change-notification tool from the sibling runtime_sandboxes_files_read/write tools without opening the schema.

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?

It clearly delineates when each action applies: start returns a watch plus cursor, read consumes the cursor, list enumerates watches, stop terminates one. It also notes access events are opt-in and that watches cap at 4 per sandbox. It never names an alternative tool or an explicit 'use this instead of X' condition, so it stops short of a 5.

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