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

Annotations are minimal (readOnlyHint=false, destructiveHint=false) and the description carries the burden. It explicitly states it runs 'REAL' commands with possible side effects, and discloses the failure mode when no active sandbox exists. This goes beyond the annotations by alerting the agent to the real execution nature and the requirement of an active 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?

The description is dense and packed with essential information in a single sentence, but it is slightly lengthy. It front-loads the core purpose ('Run a REAL command') and then lists the kind-specific parameters. The grammar is tight, though the parenthetical list of parameters could be more readable.

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 the high complexity (10 params, 3 kinds, nested objects, no output schema), the description covers the main paths but leaves gaps. It doesn't explain return values, error cases (beyond no active sandbox), or how `arguments` is structured. It is adequate for an agent to start using it but not fully complete for edge cases like the `image` parameter.

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 0%, so the description must compensate. It explains that kind can be 'bash' with `command`, 'mcp' with `endpoint`, `action`, `tool`, `arguments`, and 'agent' with `instruction`, which gives meaning to those parameters. However, it does not explain parameters like `image`, `params`, or details about `endpoint` syntax, leaving some ambiguity.

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 'Run a REAL command' in the user's active session sandbox, specifies three kinds (bash, mcp, agent) with their required parameters, and uses explicit action words. It distinguishes itself from siblings with names like sandbox_start/status/stop, which manage the sandbox, while sandbox_exec executes inside it.

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 implies usage for executing commands in the active sandbox, mentions 'No active sandbox → says so' as a precondition, and implies it is the tool to use for running commands as opposed to managing the sandbox. It does not explicitly state when NOT to use it or mention alternatives like sandbox_start for creating a sandbox, but the context is clear enough.

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